R/01-plotMap.R
In Pandora-IsoMemo/iso-app: Pandora & IsoMemo spatiotemporal modeling
Defines functions
getPColor
selectClusterGrouping
getMinima
createDifferenceMap
combineSimilarityMaps
similarityMap
createSimilarityMap
roundUpNice
plotTimeIntervals
addFormattedAxis
plotTimeCourse
north.arrow
getPlotRegion
filled.contour2
plotDS
plotMap3D
plotMap
Documented in
addFormattedAxis
filled.contour2
getPColor
plotDS
plotMap
plotMap3D
plotTimeCourse
selectClusterGrouping
#' Plots map of a spatial average or spread model from estimateMap() or estimateMapSpread() functions
#'
#' @param model return object of spatial or spread model
#' from estimateMap() or estimateMapSpread() functions
#' @param IndSelect for categorical model: selected category; shifts between categories in the center
#' @param arrow display north arrow TRUE/FALSE
#' @param scale display scale TRUE/FALSE
#' @param terrestrial show only estimates on land masses (1), oceans (-1) or all (0)
#' @param resolution spatial grid resolution of displayed (higher is slower but better quality)
#' @param interior show only convex hull TRUE/FALSE
#' @param grid show coordinate grid TRUE/FALSE
#' @param ncol number of colors for estimates
#' @param colors color scheme of estimates from RColorBrewer. defaults to "RdYlGn"
#' @param reverseColors inverse colour scheme
#' @param mapType Type of map. "Map" for simple estimates. "speed" for gradient
#' (mainly for spread model)
#' @param estType one of "Mean", "SE" and "Quantile". defaults to "Mean"
#' @param estQuantile quantile (only applicable if estType = "Quantile")
#' @param points show points on map TRUE/FALSE
#' @param pColor color of location marks drawn
#' @param colorsP color scale of location marks
#' @param pointShape pch shape of location marks
#' @param pointSize point size (if points = TRUE)
#' @param centerMap center of map, one of "Europe" and "Pacific"
#' @param StdErr maximum standard error to which estimates are displayed
#' @param rangex range of longitude values (x axis limits)
#' @param rangey range of latitude values (y axis limits)
#' @param mask boolean: Show output within range of points
#' @param maskRadius numeric: Show output within range of points in km
#' @param limitz restrict range of z (possible values "0-1", "0-100")
#' @param rangez range of estimated values (z axis limits)
#' @param dataCenter optional data.frame with two columns (longitude / latitude) for batch estimates
#' of a series of spatial points
#' @param RadiusBatch radius of batch spatial point estimates
#' @param textLabels text labels
#' @param pointLabels point size labels
#' @param pointColLabels point colour labels
#' @param fontSize font size
#' @param showScale show colour scale
#' @param setAxisLabels show axis/main/scale labels
#' @param mainLabel main label
#' @param yLabel y lab label
#' @param xLabel y lab label
#' @param scLabel scale lab label
#' @param northSize size of north arrow
#' @param scalSize size of scale
#' @param scaleX scale x orientation
#' @param scaleY scale y orientation
#' @param NorthX north arrow x orientation
#' @param NorthY north arrow y orientation
#' @param showModel show model
#' @param fontType font type
#' @param fontCol font color
#' @param titleMain show main title
#' @param titleScale show scale title
#' @param AxisSize axis title font size
#' @param AxisLSize axis label font size
#' @param cluster show clusters
#' @param clusterCol Cluster colors
#' @param pointDat data frame of points to add to plot
#' @param MinMax Min or Max, only for spread model
#' @param nMin Number of minima to compare, only for spread model
#' @param minDist Distance between minima/maxima, only for spread model
#' @param showMinOnMap Show minima on map yes/no, only for spread model if maptype = "Minima/Maxima"
#' @inheritParams filled.contour2
#'
#' @export
plotMap <- function(model,
IndSelect = NULL,
estType = "Mean", estQuantile = 0.9,
points = TRUE, pointSize = 1, StdErr = Inf,
rangex = range(model$data$Longitude, na.rm = TRUE),
rangey = range(model$data$Latitude, na.rm = TRUE),
rangez = NULL,
contourType = c("filled.contour", "contour"),
limitz = NULL, resolution = 100, interior = TRUE,
ncol = 10, colors = "RdYlGn", reverseColors = FALSE,
colorsP = NULL,
centerMap = "Europe",
pointShape = 4,
showScale = TRUE,
textLabels = NULL,
pointLabels = NULL,
pointColLabels = NULL,
fontSize = 11,
showModel = TRUE,
fontType = "sans",
fontCol = "black",
mask = FALSE,
maskRadius = 100,
pColor = "#000000",
dataCenter = NULL, RadiusBatch = NULL, terrestrial = 1,
grid = TRUE, arrow = TRUE, scale = TRUE, mapType = "Map",
titleMain = TRUE,
titleScale = TRUE,
setAxisLabels = FALSE,
mainLabel = "",
yLabel = "Latitude",
xLabel = "Longitude",
northSize = 0.2,
scalSize = 0.1,
scaleX = 0,
scaleY = 0.1,
NorthX = 0.025,
NorthY = 0.925,
scLabel = "",
AxisSize = 1,
AxisLSize = 1,
cluster = FALSE,
clusterCol = "Set1",
pointDat = NULL,
MinMax = "Min",
nMin = 3,
minDist = 250,
showMinOnMap = FALSE){
options(scipen=999)
contourType <- match.arg(contourType)
minRangeFactor <- 0.75
if((diff(rangex) / diff(rangey)) < minRangeFactor){
rangex[1] <- max(-180, mean(rangex) - minRangeFactor / 2 * diff(rangey))
rangex[2] <- min(180, mean(rangex) + minRangeFactor / 2 * diff(rangey))
}
minRangeFactor <- 0.5
if((diff(rangey) / diff(rangex)) < minRangeFactor){
rangey[1] <- max(-90, mean(rangey) - minRangeFactor / 2 * diff(rangex))
rangey[2] <- min(90, mean(rangey) + minRangeFactor / 2 * diff(rangex))
}
independent <- model$independent
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
if(IndSelect == "" | is.null(IndSelect)){
return(NULL)
}
model$model <- model$model[[IndSelect]]
}
if(is.null(rangez)){
rangez = range(model$data[, independent], na.rm = TRUE)
}
Bayes = TRUE
GAM = FALSE
if("gamm" %in% class(model$model)){
Bayes <- FALSE
}
if("kde" %in% class(model$model)){
GAM <- TRUE
independent <- "Density"
}
Maps <- loadMaps()
data <- model$data
if(!is.null(textLabels)){
textLabels <- textLabels[as.numeric(rownames(data)),1]
}
if(!is.null(pointLabels)){
pointLabels <- pointLabels[as.numeric(rownames(data)),1]
}
if(!is.null(pointColLabels)){
pointColLabels <- pointColLabels[as.numeric(rownames(data)),1]
}
sc <- model$sc
if (is.null(data)) return(NULL)
RadiusBatch <- RadiusBatch / 111
maskRadius <- maskRadius / 111
longitudes <- seq(max(-180, rangex[1] - 0.1 * diff(rangex)),
min(180, rangex[2] + 0.1 * diff(rangex)), length.out = resolution)
latitudes <- seq(max(-90, rangey[1] - 0.1 * diff(rangey)),
min(90, rangey[2] + 0.1 * diff(rangey)), length.out = resolution)
XPred <- expand.grid(Longitude = longitudes, Latitude = latitudes)
if(centerMap != "Europe"){
rangexEU <- rangex
rangex[rangexEU < 20] <- rangex[rangexEU < 20] + 160
rangex[rangexEU >= 20] <- (- 200 + rangex[rangexEU >= 20])
rangex <- rangex[order(rangex)]
longitudesPac <- longitudes
longitudes[longitudesPac < 20] <- longitudes[longitudesPac < 20] + 160
longitudes[longitudesPac >= 20] <- (- 200 + longitudes[longitudesPac >= 20])
XPredPac <- XPred
XPred$Longitude[XPredPac$Longitude < 20] <- XPred$Longitude[XPredPac$Longitude < 20] + 160
XPred$Longitude[XPredPac$Longitude >= 20] <- (- 200 + XPred$Longitude[XPredPac$Longitude >= 20])
dataPac <- data
dataPac$Longitude[data$Longitude < -20] <- dataPac$Longitude[data$Longitude < -20] + 200
dataPac$Longitude[data$Longitude >= -20] <- (- 160 + dataPac$Longitude[data$Longitude >= -20])
}
if (interior == TRUE){
if(centerMap != "Europe"){
poly.pnts <-
cbind(dataPac[chull(x = dataPac$Longitude, y = dataPac$Latitude), "Longitude"],
dataPac[chull(x = dataPac$Longitude, y = dataPac$Latitude), "Latitude"])
poly.pnts <- rbind(poly.pnts, poly.pnts[1, ])
draw <- point.in.polygon(XPredPac[, 1], XPredPac[, 2], poly.pnts[, 1], poly.pnts[, 2])
} else {
poly.pnts <-
cbind(data[chull(x = data$Longitude, y = data$Latitude), "Longitude"],
data[chull(x = data$Longitude, y = data$Latitude), "Latitude"])
poly.pnts <- rbind(poly.pnts, poly.pnts[1, ])
draw <- point.in.polygon(XPred[, 1], XPred[, 2], poly.pnts[, 1], poly.pnts[, 2])
}
}
if(mask == TRUE){
maskDraw <- extractMaskDraw(dat = centerPlotData(data, centerMap = centerMap),
maskRadius = maskRadius,
XPred = centerXPred(XPred = XPred,
XPredPac = XPredPac,
centerMap = centerMap))
}
if (!is.null(dataCenter)){
# this is batch mode
XPred <- do.call("rbind", lapply(1:nrow(dataCenter), function(x) {
longitudes <- seq(dataCenter[x, 1] - RadiusBatch,
dataCenter[x, 1] + RadiusBatch, length.out = 4)
latitudes <- seq(dataCenter[x, 2] - RadiusBatch,
dataCenter[x, 2] + RadiusBatch, length.out = 4)
XPred <- expand.grid(Longitude = longitudes, Latitude = latitudes)
cbind(XPred, id = x)
}))
}
XPred <- centerData(XPred, center = centerMap)
if (Bayes == TRUE & GAM == FALSE){
PredMatr <- Predict.matrix(sc, data = XPred)
betas <- model$model$beta
betaSigma <- model$model$betaSigma
Predictions <-
sapply(1:nrow(betas), function(x)
(PredMatr %*% betas[x, ]) * model$sRe + model$mRe)
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
Predictions <- invLogit(Predictions)
}
if(!is.null(betaSigma)){
PredMatrV <- Predict.matrix(model$scV, data = XPred)
PredictionsSigma <-
rowMeans(sqrt(sapply(1:nrow(betaSigma), function(x)
exp((PredMatrV %*% betaSigma[x, ])) / model$model$sigma[x]) * model$sRe^2))
} else {
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
PredictionsSigma <- sqrt(Predictions * (1-Predictions))
} else {
PredictionsSigma <- sqrt(mean(model$model$sigma) * model$sRe^2)
}
}
if(estType == "Mean"){
Est <- rowMeans(Predictions)
}
if(estType == "1 SE"){
Est <- apply(Predictions, 1, sd)
}
if(estType == "2 SE"){
Est <- apply(Predictions, 1, sd) * 2
}
if(estType == "1 SETOTAL"){
Est <- sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2)
}
if(estType == "2 SETOTAL"){
Est <- sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2) * 2
}
if(estType == "1 SD Population"){
Est <- PredictionsSigma * 1
}
if(estType == "2 SD Population"){
Est <- PredictionsSigma * 2
}
if(estType == "Quantile"){
Est <- apply(Predictions, 1, quantile, estQuantile)
}
if(estType == "QuantileTOTAL"){
Est <- rowMeans(Predictions + qnorm(estQuantile) * sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2))
}
XPred <- data.frame(XPred,
Est = Est,
Sd = apply(Predictions, 1, sd),
SDPop = PredictionsSigma,
SdTotal = sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2),
IntLower = apply(Predictions, 1, quantile, 0.025),
IntUpper = apply(Predictions, 1, quantile, 0.975),
IntLowerTotal = apply(Predictions + sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2),
1, quantile, 0.025),
IntUpperTotal = apply(Predictions + sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2),
1, quantile, 0.975),
resError = sqrt(mean(model$model$sigma + model$model$tau) * model$sRe^2))
}
if (Bayes == FALSE & GAM == FALSE){
Est <- predict(model$model$gam, XPred, se.fit = TRUE, type = "response")
if(estType == "1 SE"){
Est$fit <- Est$se.fit
}
if(estType == "2 SE"){
Est$fit <- Est$se.fit * 2
}
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
varM = Est$fit * (1-Est$fit)
} else {
varM = var(residuals(model$model$gam))
}
if(estType == "1 SETOTAL"){
Est$fit <- sqrt(Est$se.fit^2 + varM)
}
if(estType == "2 SETOTAL"){
Est$fit <- sqrt(Est$se.fit^2 + varM) * 2
}
if(estType == "1 SD Population"){
Est$fit <- sqrt(varM) * 1
}
if(estType == "2 SD Population"){
Est$fit <- sqrt(varM) * 2
}
if(estType == "Quantile"){
Est$fit <- Est$fit + qnorm(estQuantile) * Est$se.fit
}
if(estType == "QuantileTOTAL"){
Est$fit <- Est$fit + qnorm(estQuantile) *
sqrt(Est$se.fit^2 + varM)
}
XPred <- data.frame(XPred,
Est = Est$fit,
Sd = Est$se.fit,
SDPop = sqrt(varM),
SdTotal = sqrt(Est$se.fit^2 + varM),
IntLower = Est$fit - 1.96 * Est$se.fit,
IntUpper = Est$fit + 1.96 * Est$se.fit,
IntLowerTotal = Est$fit - 1.96 * sqrt(Est$se.fit^2 + varM),
IntUpperTotal = Est$fit + 1.96 * sqrt(Est$se.fit^2 + varM),
resError = sqrt(mean((predict(model$model$gam, type = "response") - model$model$gam$y)^2)))
}
if(GAM == TRUE){
Predictions <- sapply(1:length(model$model), function(x) predict(model$model[[x]], x = XPred[, 1:2]))
if(estType == "Mean"){
Est <- rowMeans(Predictions)
}
if(estType == "1 SE"){
Est <- apply(Predictions, 1, sd)
}
if(estType == "2 SE"){
Est <- apply(Predictions, 1, sd) * 2
}
if(estType == "Quantile"){
Est <- apply(Predictions, 1, quantile, estQuantile)
}
XPred <- data.frame(XPred,
Est = Est,
Sd = apply(Predictions, 1, sd),
IntLower = pmax(0, apply(Predictions, 1, quantile, 0.025)),
IntUpper = apply(Predictions, 1, quantile, 0.975))
}
if (estType != "1 SE" && estType != "2 SE" &&
estType != "1 SETOTAL" &&
estType != "2 SETOTAL" &&
!is.null(limitz) && !missing(limitz)){
if (limitz == "0-1"){
XPred$Est <- pmin(1, XPred$Est)
XPred$Est <- pmax(0, XPred$Est)
if(!is.null(XPred$IntLower)){
XPred$IntLower <- pmax(0, pmin(1, XPred$IntLower))
XPred$IntUpper <- pmax(0, pmin(1, XPred$IntUpper))
}
if(!is.null(XPred$IntLowerTotal)){
XPred$IntLowerTotal <- pmax(0, pmin(1, XPred$IntLowerTotal))
XPred$IntUpperTotal <- pmax(0, pmin(1, XPred$IntUpperTotal))
}
}
if (limitz == "0-100"){
XPred$Est <- pmin(100, XPred$Est)
XPred$Est <- pmax(0, XPred$Est)
if(!is.null(XPred$IntLower)){
XPred$IntLower <- pmax(0, pmin(100, XPred$IntLower))
XPred$IntUpper <- pmax(0, pmin(100, XPred$IntUpper))
}
if(!is.null(XPred$IntLowerTotal)){
XPred$IntLowerTotal <- pmax(0, pmin(100, XPred$IntLowerTotal))
XPred$IntUpperTotal <- pmax(0, pmin(100, XPred$IntUpperTotal))
}
}
}
if (!is.null(dataCenter)){
# this is batch mode
XPredCenter <- lapply(1:nrow(dataCenter), function(x){
XPred %>%
extractXPredCenter(centerX = dataCenter[x, 1],
centerY = dataCenter[x, 2],
Radius = RadiusBatch)
})
centerEstimates <- lapply(1:nrow(dataCenter), function(x){
XPredCenter[[x]] %>%
extractCenterEstimates(batch = TRUE)
})
meanCenter <- sapply(1:nrow(dataCenter), function(x) centerEstimates[[x]]$mean)
sdCenter <- sapply(1:nrow(dataCenter), function(x) centerEstimates[[x]]$sd)
IntLower <- sapply(1:nrow(dataCenter),
function(x) signif(mean(XPredCenter[[x]]$IntLower), 5))
IntUpper <- sapply(1:nrow(dataCenter),
function(x) signif(mean(XPredCenter[[x]]$IntUpper), 5))
if(is.null(XPredCenter[[1]]$SDPop)) {
dataCenter <- cbind(dataCenter, mean = meanCenter, sd = sdCenter,
IntLower = IntLower, IntUpper = IntUpper)
} else {
#population SD
SDPop <- sapply(1:nrow(dataCenter),
function(x) signif(sqrt(mean(XPredCenter[[x]]$SDPop, na.rm = TRUE)^2), 5))
sdCenterTotal <- sapply(1:nrow(dataCenter),
function(x) signif(sqrt(sum(sd(XPredCenter[[x]]$Est) ^ 2,
mean(XPredCenter[[x]]$Sd) ^ 2,
mean(XPredCenter[[x]]$SDPop) ^ 2,
na.rm = TRUE)), 5))
IntLowerTotal <- sapply(1:nrow(dataCenter),
function(x) signif(min(XPredCenter[[x]]$IntLowerTotal), 5))
IntUpperTotal <- sapply(1:nrow(dataCenter),
function(x) signif(max(XPredCenter[[x]]$IntUpperTotal), 5))
dataCenter <- cbind(dataCenter, mean = meanCenter, sd = sdCenter,
SDPop = SDPop,
sdTotal = sdCenterTotal,
IntLower = IntLower, IntUpper = IntUpper,
IntLowerTotal = IntLowerTotal, IntUpperTotal = IntUpperTotal)
}
return(dataCenter)
}
# keep $Est in XPred for later calculation of mean and sd
XPred$EstForCenter <- XPred$Est
if (interior == TRUE){
# this step can remove all $Est
XPred$Est[draw == 0] <- NA
}
if (mask == TRUE){
XPred$Est[maskDraw == 0] <- NA
}
if(GAM == TRUE){
XPred$Est[is.na(XPred$Est) | XPred$Est < 0] <- 0
}
XPred$Est[XPred$Sd > StdErr] <- NA
levels <- pretty(c(rangez[1], rangez[2]), n = ncol)
if(mapType == "Minima/Maxima"){
mins <- getMinima(XPred, nMin = nMin, minDist = minDist, minima = MinMax)
if(MinMax == "Min"){
probs <- round(as.vector(prop.table(table(factor(apply(Predictions[mins, ], 2, which.min), levels = 1:nMin)))), 3)
} else{
probs <- round(as.vector(prop.table(table(factor(apply(Predictions[mins, ], 2, which.max), levels = 1:nMin)))), 3)
}
dataMinPlot <- XPred[mins,]
dataMinPlotPred <- (t(Predictions[mins,]))
dataMinPlotPred <- data.frame(name = rep(paste0(MinMax, "ima ", 1:nMin,
"\n Coord. = (",round(dataMinPlot$Latitude, 2),", ",
round(dataMinPlot$Longitude, 2),") \n Probability = ", prob = probs),
each = nrow(dataMinPlotPred)),
estimate = as.vector(dataMinPlotPred))
minMeans <- aggregate(dataMinPlotPred$estimate, by=list(name=dataMinPlotPred$name), FUN=mean)
minMedians <- aggregate(dataMinPlotPred$estimate, by=list(name=dataMinPlotPred$name), FUN=median)
minq68 <- aggregate(dataMinPlotPred$estimate, by=list(name=dataMinPlotPred$name), FUN=quantile, 0.68)
minq32 <- aggregate(dataMinPlotPred$estimate, by=list(name=dataMinPlotPred$name), FUN=quantile, 0.32)
minq95 <- aggregate(dataMinPlotPred$estimate, by=list(name=dataMinPlotPred$name), FUN=quantile, 0.975)
minq05 <- aggregate(dataMinPlotPred$estimate, by=list(name=dataMinPlotPred$name), FUN=quantile, 0.025)
meanEst <- q32 <- q68 <- q95 <- q05 <- NULL
dataSummary <- data.frame(name = unique(dataMinPlotPred$name), meanEst = minMeans[,2], median = minMedians[,2],
q68 = minq68[,2], q32 = minq32[,2],
q95 = minq95[,2], q05 = minq05[,2])
if(showMinOnMap == "1"){
g <- ggplot(dataSummary, aes_(x = ~name, fill = ~name)) + theme_light()
g <- g + geom_boxplot(
mapping = aes(
lower = q32,
upper = q68,
middle = median,
ymin = q05,
ymax = q95
),
stat = "identity"
) + geom_errorbar(aes(ymin = meanEst, ymax = meanEst), linetype = "dashed", data = dataSummary)+ theme(legend.position="none") +
labs(title=paste0("Comparison of local ", MinMax, "ima")) + xlab("")
print(g)
return(list(XPred = XPred))
}
}
#if (!all(is.na(XPred$Est))){
if (mapType == "Speed"){
if(centerMap != "Europe"){
longitudes <- longitudesPac[order(longitudesPac)]
latitudes <- latitudes[order(latitudes, longitudesPac)]
XPredPlot <- data.frame(XPredPac[order(XPredPac$Latitude, XPredPac$Longitude),],
Est = XPred$Est[order(XPredPac$Latitude, XPredPac$Longitude)])
rangex <- rangexEU
} else {
XPredPlot <- XPred
}
z = matrix(XPredPlot$Est, ncol = resolution)
test.rast <- raster(ncol= nrow(z), nrow = nrow(z), xmn = 0, xmx = 1, ymn = 0, ymx = 1)
test.rast[] <- as.vector(z)
z2 <- abs(matrix(getValues(terrain(test.rast, unit = "tangent")), ncol = nrow(z)))
tankilometers <- sqrt(diff(latitudes)[1] * diff(longitudes)[1]) * 111
z2 <- log(((1/z2) / tankilometers) + 1E-10)
z2levels <- signif(exp(pretty(as.vector(z2), n = ncol)), 2)
q99 <- quantile(as.vector(exp(z2)), na.rm = TRUE, 0.975)
z2levels <- z2levels[z2levels < q99]
z2levels <- unique(c(z2levels, signif(q99, 2),
signif(max(as.vector(exp(z2)), na.rm = TRUE), 2)))
# z2levels <- signif(pretty(c((rangez[1])^0.05, (rangez[2])^0.05), n = ncol)^20, 2)
# z2levels[1] <- rangez[1]
# z2levels[length(z2levels)] <- rangez[2]
# if(length(unique(z2levels)) < 4){
# z2levels <- signif(pretty(c((1)^0.05, (1000)^0.05), n = ncol)^20, 2)
# }
colors <- colorRampPalette(brewer.pal(9, colors))(length(z2levels) - 1)
if(reverseColors){
colors <- rev(colors)
}
levelsLegend <- z2levels
if(!showModel){
z2 <- rep(NA, length(z2))
}
if(titleMain){
main = ""
} else {
if(setAxisLabels){
main = mainLabel
} else {
main = independent
}
}
if(titleScale){
mainS = ""
} else {
if(setAxisLabels){
mainS = scLabel
} else {
mainS = "km/time unit"
}
}
if(setAxisLabels){
xlab = xLabel
ylab = yLabel
} else {
xlab = "Longitude"
ylab = "Latitude"
}
#for geotiff export
if(centerMap != "Europe"){
XPred$Est <- exp(as.vector(z2))[order(XPred$Latitude, XPred$Longitude)]
} else {
XPred$Est <- exp(as.vector(z2))
}
filled.contour2(longitudes, latitudes, z = z2,
contourType = contourType,
cex.axis = 1.5, cex.main = 1.5, cex.lab = 1.5,
xlim = rangex, ylim = rangey,
zlim = range(z2, finite = TRUE),
levels = log(z2levels + 1E-10), nlevels = length(z2levels),
col = colors,
asp = 1,
showScale = showScale,
key.axes = axis(4, at = log(levelsLegend + 1E-10), labels = levelsLegend),
key.title = title(main = mainS, cex.main = 0.8),
plot.title = {title(cex.lab = AxisSize, xlab = xlab, ylab = ylab, main = main)},
plot.axes = {
par(fg = "black", col="black");
if (terrestrial == 1){
if(centerMap != "Europe"){
sp::plot(Maps$ocean160, add = T, col = "lightblue", lwd = 1, border = NA);
sp::plot(Maps$ocean200, add = T, col = "lightblue", lwd = 1, border = NA);
} else {
sp::plot(Maps$ocean, add = T, col = "lightblue", lwd = 1);
}
}
if (terrestrial == -1){
if(centerMap != "Europe"){
sp::plot(Maps$land160, add = T, lwd = 1, col = "grey96", border = NA);
sp::plot(Maps$land200, add = T, lwd = 1, col = "grey96", border = NA);
} else {
sp::plot(Maps$land, add = T, lwd = 1, col = "grey96", border = NA);
}
}
if(centerMap != "Europe"){
sp::plot(Maps$coast160, add = T, lwd = 1);
sp::plot(Maps$coast200, add = T, lwd = 1);
} else {
sp::plot(Maps$coast, add = T, lwd = 1);
}
if (grid == TRUE){
if(centerMap != "Europe"){
sp::plot(Maps$grids160, add = T, col = "grey", lty = 2, xlim = c(0, 1));
sp::plot(Maps$grids200, add = T, col = "grey", lty = 2, xlim = c(0, 1));
} else {
sp::plot(Maps$grids, add = T, col = "grey", lty = 2, xlim = c(0, 1));
}
}
if(centerMap != "Europe"){
sp::plot(Maps$borders160, add = T, col = "darkgrey", lwd = 1);
sp::plot(Maps$borders200, add = T, col = "darkgrey", lwd = 1);
} else {
sp::plot(Maps$borders, add = T, col = "darkgrey", lwd = 1);
}
if (points == TRUE){
if(!is.null(pointLabels)){
pointLabels <- as.numeric(pointLabels)
pointLabels <- pointLabels - min(pointLabels, na.rm = TRUE)
pointLabels <- pointLabels / max(pointLabels, na.rm = TRUE)
pointSize <- (pointSize * pointLabels + 0.1) * 2
}
if(!is.null(pointColLabels)){
pointColLabels <- as.numeric(pointColLabels)
pointColLabels <- pointColLabels - min(pointColLabels, na.rm = TRUE)
pointColLabels <- pointColLabels / max(pointColLabels, na.rm = TRUE)
pColor <- (colorRampPalette(brewer.pal(9, colorsP))(101))[round(pointColLabels, 2) * 100]
}
if(cluster & !is.null(data$spatial_cluster)){
pColor <- colorRampPalette(brewer.pal(8, clusterCol))(max(data$spatial_cluster, na.rm=TRUE))[data$spatial_cluster]
if((!"long_temporal_group_reference_point" %in% names(data))){
data_names <- c("spatial_cluster", "long_centroid_spatial_cluster", "lat_centroid_spatial_cluster")
} else {
data_names <- c("temporal_group", "long_temporal_group_reference_point", "lat_temporal_group_reference_point")
}
centroids <- unique(data[, data_names])
centroids <- centroids[order(centroids[,1]), ]
if(centerMap != "Europe"){
centroids2 <- centroids
centroids2[,2][centroids[,2] < -20] <- centroids2[,2][centroids[,2] < -20] + 200
centroids2[,2][centroids[,2] >= -20] <- (- 160 + centroids2[,2][centroids[,2] >= -20])
centroids <- centroids2
}
}
if(centerMap != "Europe"){
points(dataPac$Latitude ~ dataPac$Longitude,
col = pColor, lwd = 2,
pch = pointShape, cex = pointSize);
} else {
points(data$Latitude ~ data$Longitude,
col = pColor, lwd = 2,
pch = pointShape, cex = pointSize);
}
if(cluster & !is.null(data$spatial_cluster)){
points(centroids[, 2:3], lwd = 2,
pch = pointShape, cex = pointSize * 2.5, col = colorRampPalette(brewer.pal(8, clusterCol))(max(data$spatial_cluster, na.rm=TRUE)))
text(centroids[, 2:3], labels = paste0("Cluster_", centroids[,1]), pos = 4,
cex = fontSize * 1.5, col = fontCol, family = fontType)
}
}
if(!is.null(textLabels)){
if(centerMap != "Europe"){
text(dataPac$Latitude ~ dataPac$Longitude,
labels = as.character(textLabels), pos = 4,
cex = fontSize, col = fontCol, family = fontType)
} else {
text(data$Latitude ~ data$Longitude,
labels = as.character(textLabels), pos = 4,
cex = fontSize, col = fontCol, family = fontType)
}
}
if(centerMap != "Europe"){
lab <- pretty(rangex)
lab[pretty(rangex) >= 20] <- lab[pretty(rangex) >= 20] - 200
lab[pretty(rangex) < 20] <- lab[pretty(rangex) < 20] + 160
axis(1, at = pretty(rangex), labels = lab, cex.axis = AxisLSize);
} else{
axis(1, cex.axis = AxisLSize);
}
axis(2, cex.axis = AxisLSize);
}
)
if (arrow == TRUE){
if(showScale == TRUE){
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
} else {
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
}
}
if (scale == TRUE){
if(showScale == TRUE){
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
} else {
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
}
}
return(list(XPred = XPred))
}
if(GAM == TRUE){
colors <- c("#FFFFFF", colorRampPalette(brewer.pal(9, colors))(length(levels) - 2))
} else{
colors <- colorRampPalette(brewer.pal(9, colors))(length(levels) - 1)
}
if(reverseColors){
colors <- rev(colors)
}
levelsLegend <- levels
if(length(levels) > 25){
par(fg = NA, col="black")
levelsLegend <- pretty(c(rangez[1], rangez[2]),
n = pmin(20, ceiling(ncol / 2)))
}
cex4 <- 1
if(max(abs(XPred$Est), na.rm = TRUE) > 9999 | max(abs(XPred$Est), na.rm = TRUE) < 0.05){
cex4 <- 0.7
}
if(centerMap != "Europe"){
longitudes <- longitudesPac[order(longitudesPac)]
latitudes <- latitudes[order(latitudes, longitudesPac)]
XPredPlot <- data.frame(XPredPac[order(XPredPac$Latitude, XPredPac$Longitude),],
Est = XPred$Est[order(XPredPac$Latitude, XPredPac$Longitude)])
rangex <- rangexEU
} else {
XPredPlot <- XPred
}
if(!showModel){
XPredPlot$Est <- rep(NA, length(XPredPlot$Est))
}
if(titleMain){
main = ""
} else {
if(setAxisLabels){
main = mainLabel
} else {
main = independent
}
}
if(titleScale){
mainS = ""
} else {
if(setAxisLabels){
mainS = scLabel
} else {
mainS = independent
}
}
if(setAxisLabels){
xlab = xLabel
ylab = yLabel
} else {
xlab = "Longitude"
ylab = "Latitude"
}
filled.contour2(longitudes, latitudes, z = matrix(XPredPlot$Est, ncol = resolution),
contourType = contourType,
xlim = rangex, ylim = rangey, levels = levels, zlim = rangez,
col = colors,
showScale = showScale,
cex.axis = 1.5, cex.main = 1.5, cex.lab = 1.5,
asp = 1, key.axes = axis(side = 4, at = levelsLegend, tick = FALSE, cex.axis = cex4),
key.title = title(main = mainS, cex.main = 0.8),
plot.title = {title(cex.lab = AxisSize, xlab = xlab, ylab = ylab, main = main)},
plot.axes = {
par(fg = "black", col="black");
if (terrestrial == 1){
if(centerMap != "Europe"){
sp::plot(Maps$ocean160, add = T, col = "lightblue", lwd = 1, border = NA);
sp::plot(Maps$ocean200, add = T, col = "lightblue", lwd = 1, border = NA);
} else {
sp::plot(Maps$ocean, add = T, col = "lightblue", lwd = 1);
}
}
if (terrestrial == -1){
if(centerMap != "Europe"){
sp::plot(Maps$land160, add = T, lwd = 1, col = "grey96", border = NA);
sp::plot(Maps$land200, add = T, lwd = 1, col = "grey96", border = NA);
} else {
sp::plot(Maps$land, add = T, lwd = 1, col = "grey96", border = NA);
}
}
if(centerMap != "Europe"){
sp::plot(Maps$coast160, add = T, lwd = 1);
sp::plot(Maps$coast200, add = T, lwd = 1);
} else {
sp::plot(Maps$coast, add = T, lwd = 1);
}
if (grid == TRUE){
if(centerMap != "Europe"){
sp::plot(Maps$grids160, add = T, col = "grey", lty = 2, xlim = c(0, 1));
sp::plot(Maps$grids200, add = T, col = "grey", lty = 2, xlim = c(0, 1));
} else {
sp::plot(Maps$grids, add = T, col = "grey", lty = 2, xlim = c(0, 1));
}
}
if(centerMap != "Europe"){
sp::plot(Maps$borders160, add = T, col = "darkgrey", lwd = 1);
sp::plot(Maps$borders200, add = T, col = "darkgrey", lwd = 1);
} else {
sp::plot(Maps$borders, add = T, col = "darkgrey", lwd = 1);
}
if (points == TRUE){
if(!is.null(pointLabels)){
pointLabels <- as.numeric(pointLabels)
pointLabels <- pointLabels - min(pointLabels, na.rm = TRUE)
pointLabels <- pointLabels / max(pointLabels, na.rm = TRUE)
pointSize <- (pointSize * pointLabels + 0.1) * 2
}
if(!is.null(pointColLabels)){
pointColLabels <- as.numeric(pointColLabels)
pointColLabels <- pointColLabels - min(pointColLabels, na.rm = TRUE)
pointColLabels <- pointColLabels / max(pointColLabels, na.rm = TRUE)
pColor <- (colorRampPalette(brewer.pal(9, colorsP))(101))[round(pointColLabels, 2) * 100]
}
if(cluster & !is.null(data$spatial_cluster)){
pColor <- colorRampPalette(brewer.pal(8, clusterCol))(max(data$spatial_cluster, na.rm=TRUE))[data$spatial_cluster]
if((!"long_temporal_group_reference_point" %in% names(data))){
data_names <- c("spatial_cluster", "long_centroid_spatial_cluster", "lat_centroid_spatial_cluster")
} else {
data_names <- c("temporal_group", "long_temporal_group_reference_point", "lat_temporal_group_reference_point")
}
centroids <- unique(data[, data_names])
centroids <- centroids[order(centroids[,1]), ]
if(centerMap != "Europe"){
centroids2 <- centroids
centroids2[,2][centroids[,2] < -20] <- centroids2[,2][centroids[,2] < -20] + 200
centroids2[,2][centroids[,2] >= -20] <- (- 160 + centroids2[,2][centroids[,2] >= -20])
centroids <- centroids2
}
}
if(centerMap != "Europe"){
points(dataPac$Latitude ~ dataPac$Longitude,
col = pColor, lwd = 2,
pch = pointShape, cex = pointSize);
} else {
points(data$Latitude ~ data$Longitude,
col = pColor, lwd = 2,
pch = pointShape, cex = pointSize);
}
if(cluster & !is.null(data$spatial_cluster)){
points(centroids[, 2:3], lwd = 2,
pch = pointShape, cex = pointSize * 2.5, col = colorRampPalette(brewer.pal(8, clusterCol))(max(data$spatial_cluster, na.rm=TRUE)))
text(centroids[, 2:3], labels = paste0("Cluster_", centroids[,1]), pos = 4,
cex = fontSize * 1.5, col = fontCol, family = fontType)
}
}
if(mapType == "Minima/Maxima"){
points(dataMinPlot$Latitude ~ dataMinPlot$Longitude, lwd = 2, col = pColor,
pch = pointShape, cex = pointSize * 2.5)
text(dataMinPlot$Latitude ~ dataMinPlot$Longitude, labels = paste0(MinMax, "ima", 1:nMin), pos = 4,
cex = fontSize * 1.5, col = fontCol, family = fontType)
}
if(!is.null(textLabels)){
if(centerMap != "Europe"){
text(dataPac$Latitude ~ dataPac$Longitude,
labels = as.character(textLabels), pos = 4,
cex = fontSize, col = fontCol, family = fontType)
} else {
text(data$Latitude ~ data$Longitude,
labels = as.character(textLabels), pos = 4,
cex = fontSize, col = fontCol, family = fontType)
}
}
if(!is.null(pointDat) & NROW(pointDat) > 0){
points(x = pointDat$x, y = pointDat$y, cex = pointDat$pointSize, col = pointDat$pointColor, pch = 16)
text(pointDat$y ~ pointDat$x, labels = pointDat$label, pos = 4, cex = 1.75)
}
if(centerMap != "Europe"){
lab <- pretty(rangex)
lab[pretty(rangex) >= 20] <- lab[pretty(rangex) >= 20] - 200
lab[pretty(rangex) < 20] <- lab[pretty(rangex) < 20] + 160
axis(1, at = pretty(rangex), labels = lab, cex.axis = AxisLSize);
} else{
axis(1, cex.axis = AxisLSize);
}
axis(2, cex.axis = AxisLSize);
}
)
# } else {
# plot(1:10, xlim = range(data$Longitude), ylim = range(data$Latitude),
# xlab = "Longitude", ylab = "Latitude")
# plot(Maps$ocean, add = T, col = "lightblue", lwd = 1)
# sp::plot(Maps$grids, add = T, col = "grey", lty = 2, xlim = c(0, 1))
# sp::plot(Maps$borders, add = T, col = "darkgrey", lwd = 1)
# text(mean(data$Longitude),mean(data$Latitude), cex = 1,
# "No estimates to plot, please adjust \"Display up to max standard error\" slider",
# col = "red")
# }
if (arrow == TRUE){
if(showScale == TRUE){
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
} else {
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
}
}
if (scale == TRUE){
if(showScale == TRUE){
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
} else {
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
}
}
return(list(XPred = XPred))
}
#' Plots time slice map of a spatio-temporal model from estimateMap3D() function
#'
#' @param model return object of a spatio-temporal model from estimateMap3D() function
#' @param IndSelect for categorical model: selected category
#' @param time time slice value for map
#' @param arrow display north arrow TRUE/FALSE
#' @param scale display scale TRUE/FALSE
#' @param terrestrial show only estimates on land masses (1), oceans (-1) or all (0)
#' @param resolution spatial grid resolution of displayed (higher is slower but better quality)
#' @param interior show only convex hull 0 "none", 1 "3D", 2 "time sliced 2D"
#' @param addU numeric: years of added uncertainty for time sliced 2D-convex hull
#' @param grid show coordinate grid TRUE/FALSE
#' @param ncol number of colors for estimates
#' @param pColor color of location marks drawn
#' @param colorsP color scale of location marks
#' @param pointShape pch shape of location marks
#' @param colors color scheme of estimates from RColorBrewer. defaults to "RdYlGn"
#' @param reverseColors inverse colour scheme
#' @param estType one of "Mean", "SE" and "Quantile". defaults to "Mean"
#' @param estQuantile quantile (only applicable if estType = "Quantile")
#' @param points show points on map TRUE/FALSE
#' @param pointSize point size (if points = TRUE)
#' @param StdErr maximum standard error to which estimates are displayed
#' @param rangex range of longitude values (x axis limits)
#' @param rangey range of latitude values (y axis limits)
#' @param mask boolean: Show output within range of points
#' @param maskRadius numeric: Show output within range of points in km
#' @param limitz restrict range of z (possible values "0-1", "0-100")
#' @param rangez range of estimated values (z axis limits)
#' @param centerMap center of map, one of "Europe" and "Pacific"
#' @param dataCenter optional data.frame with three columns (longitude / latitude / time)
#' @param RadiusBatch radius of batch spatial point estimates
#' @param textLabels text labels
#' @param pointLabels point size labels
#' @param pointColLabels point colour labels
#' @param fontSize font size
#' @param showModel show model
#' @param fontType font type
#' @param fontCol font color
#' @param titleMain show main title
#' @param titleScale show scale title
#' @param setAxisLabels show axis/main/scale labels
#' @param mainLabel main label
#' @param yLabel y lab label
#' @param xLabel y lab label
#' @param northSize size of north arrow
#' @param scalSize size of scale
#' @param scaleX scale x orientation
#' @param scaleY scale y orientation
#' @param NorthX north arrow x orientation
#' @param NorthY north arrow y orientation
#' @param scLabel scale lab label
#' @param AxisSize axis title font size
#' @param AxisLSize axis label font size
#' @param cluster show clusters
#' @param clusterAll show all cluster points
#' @param clusterResults temporal groups or spatial clusters
#' @param clusterCol Cluster colors
#' @param pointDat data frame of points to add to plot
#' @param plotRetNull return predictions
#' @inheritParams filled.contour2
#'
#' @export
plotMap3D <- function(model,
time,
IndSelect = NULL,
estType = "Mean", estQuantile = 0.9,
points = TRUE, pointSize = 1, StdErr = Inf,
rangex = range(model$data$Longitude, na.rm = TRUE),
rangey = range(model$data$Latitude, na.rm = TRUE),
rangez = NULL,
contourType = c("filled.contour", "contour"),
limitz = "none",
centerMap = "Europe",
addU = 0,
pointShape = 4,
pColor = "#000000",
colorsP = NULL,
showModel = TRUE,
textLabels = NULL,
pointLabels = NULL,
pointColLabels = NULL,
fontSize = 11,
mask = FALSE,
maskRadius = 100,
showScale = TRUE,
fontType = "sans",
fontCol = "black",
resolution = 100, interior = TRUE,
ncol = 10, colors = "RdYlGn", reverseColors = FALSE,
dataCenter = NULL,
RadiusBatch = 100, terrestrial = 1,
grid = TRUE, arrow = TRUE, scale = TRUE,
titleMain = TRUE,
titleScale = TRUE,
setAxisLabels = FALSE,
mainLabel = "",
yLabel = "Latitude",
xLabel = "Longitude",
scLabel = "",
northSize = 0.2,
scalSize = 0.1,
scaleX = 0,
scaleY = 0.1,
NorthX = 0.025,
NorthY = 0.925,
AxisSize = 1,
AxisLSize = 1,
cluster = FALSE,
clusterAll = "0",
clusterResults = 0,
clusterCol = "Set1",
pointDat = NULL,
plotRetNull = FALSE){
options(scipen = 999)
contourType <- match.arg(contourType)
minRangeFactor <- 0.75
if((diff(rangex) / diff(rangey)) < minRangeFactor){
rangex[1] <- max(-180, mean(rangex) - minRangeFactor / 2 * diff(rangey))
rangex[2] <- min(180, mean(rangex) + minRangeFactor / 2 * diff(rangey))
}
minRangeFactor <- 0.5
if((diff(rangey) / diff(rangex)) < minRangeFactor){
rangey[1] <- max(-90, mean(rangey) - minRangeFactor / 2 * diff(rangex))
rangey[2] <- min(90, mean(rangey) + minRangeFactor / 2 * diff(rangex))
}
independent <- model$independent
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
if(is.null(IndSelect) || IndSelect == ""){
return(NULL)
}
model$model <- model$model[[IndSelect]]
}
if(is.null(rangez)){
rangez = range(model$data[, independent], na.rm = TRUE)
}
Bayes = TRUE
GAM = FALSE
if("gamm" %in% class(model$model)){
Bayes = FALSE
}
if("kde" %in% class(model$model)){
GAM = TRUE
independent = "Density"
}
Maps <- loadMaps()
data <- model$data
sc <- model$sc
if(!is.null(textLabels)){
textLabels <- textLabels[as.numeric(rownames(data)),1]
}
if(!is.null(pointLabels)){
pointLabels <- pointLabels[as.numeric(rownames(data)),1]
}
if(!is.null(pointColLabels)){
pointColLabels <- pointColLabels[as.numeric(rownames(data)),1]
}
if (is.null(data)) return(NULL)
RadiusBatch <- RadiusBatch / 111
maskRadius <- maskRadius / 300
longitudes <- seq(max(-180, rangex[1] - 0.1 * diff(rangex)),
min(180, rangex[2] + 0.1 * diff(rangex)), length.out = resolution)
latitudes <- seq(max(-90, rangey[1] - 0.1 * diff(rangey)),
min(90, rangey[2] + 0.1 * diff(rangey)), length.out = resolution)
XPred <- cbind(expand.grid(Longitude2 = (longitudes - mean(data$Longitude)) / sd(data$Longitude),
Latitude2 = (latitudes - mean(data$Latitude)) / sd(data$Latitude)),
expand.grid(Longitude = longitudes, Latitude = latitudes), Date = time)
if(centerMap != "Europe"){
rangexEU <- rangex
rangex[rangexEU < 20] <- rangex[rangexEU < 20] + 160
rangex[rangexEU >= 20] <- (- 200 + rangex[rangexEU >= 20])
rangex <- rangex[order(rangex)]
longitudesPac <- longitudes
longitudes[longitudesPac < 20] <- longitudes[longitudesPac < 20] + 160
longitudes[longitudesPac >= 20] <- (- 200 + longitudes[longitudesPac >= 20])
XPredPac <- XPred
XPred$Longitude[XPredPac$Longitude < 20] <- XPred$Longitude[XPredPac$Longitude < 20] + 160
XPred$Longitude[XPredPac$Longitude >= 20] <- (- 200 + XPred$Longitude[XPredPac$Longitude >= 20])
XPred$Longitude2 = (XPred$Longitude - mean(data$Longitude)) / sd(data$Longitude)
dataPac <- data %>%
centerPlotData(centerMap = centerMap)
dataTPac <- dataPac %>%
filterT(addU = addU, time = time)
}
if (interior > 0){
if(centerMap != "Europe"){
if(interior == 1){
cData <- rbind(data.frame(Date = dataPac$Date + 2 * dataPac$Uncertainty, dataPac[, c("Longitude", "Latitude")]),
data.frame(Date = dataPac$Date - 2 * dataPac$Uncertainty, dataPac[, c("Longitude", "Latitude")]))
if(nrow(cData) > 0){
convHull <- convhulln(cData)
draw <- inhulln(convHull, as.matrix(XPredPac[, c(5,3,4)]))
} else {
draw <- rep(0, nrow(XPred))
}
} else {
cData <- dataTPac[, c("Longitude", "Latitude")]
if(nrow(cData) > 2){
convHull <- convhulln(cData)
draw <- inhulln(convHull, as.matrix(XPredPac[, c(3,4)]))
} else {
draw <- rep(0, nrow(XPred))
}
}
} else {
if(interior == 1){
cData <- rbind(data.frame(Date = data$Date + 2 * data$Uncertainty, data[, c("Longitude", "Latitude")]),
data.frame(Date = data$Date - 2 * data$Uncertainty, data[, c("Longitude", "Latitude")]))
if(nrow(unique(cData)) > 3){
convHull <- convhulln(cData)
draw <- inhulln(convHull, as.matrix(XPred[, c(5,3,4)]))
} else {
draw <- rep(0, nrow(XPred))
}
} else {
cData <- filterT(data, addU = addU, time = time)[, c("Longitude", "Latitude")]
if(nrow(unique(cData)) > 3){
convHull <- convhulln(cData)
draw <- inhulln(convHull, as.matrix(XPred[, c(3,4)]))
} else {
draw <- rep(0, nrow(XPred))
}
}
}
}
if(mask == TRUE){
if(exists("cData")){
maskData <- unique(cData[, c("Longitude", "Latitude")])
} else {
maskData <- data
}
maskDraw <- extractMaskDraw(dat = maskData,
maskRadius = maskRadius,
XPred = centerXPred(XPred = XPred,
XPredPac = XPredPac,
centerMap = centerMap))
}
if (!is.null(dataCenter)){
# this is batch mode
XPred <- do.call("rbind", lapply(1:nrow(dataCenter), function(x) {
longitudes <- seq(dataCenter[x, 1] - RadiusBatch,
dataCenter[x, 1] + RadiusBatch, length.out = 4)
latitudes <- seq(dataCenter[x, 2] - RadiusBatch,
dataCenter[x, 2] + RadiusBatch, length.out = 4)
XPred <- expand.grid(Longitude2 = (longitudes - mean(data$Longitude)) / sd(data$Longitude),
Latitude2 = (latitudes - mean(data$Latitude)) / sd(data$Latitude),
time = dataCenter[x, 3])
XPred$Longitude <- XPred$Longitude2 * sd(data$Longitude) + mean(data$Longitude)
XPred$Latitude <- XPred$Latitude2 * sd(data$Latitude) + mean(data$Latitude)
cbind(XPred, id = x)
}))
}
if (Bayes == TRUE & GAM == FALSE){
PredMatr <- Predict.matrix(sc,
data = data.frame(XPred,
Date2 = (time - mean(data$Date)) /
sd(data$Date)))
betas <- model$model$beta
betaSigma <- model$model$betaSigma
Predictions <-
sapply(1:nrow(betas), function(x)
PredMatr %*% betas[x, ] * model$sRe + model$mRe)
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
Predictions <- invLogit(Predictions)
}
if(!is.null(betaSigma)){
PredMatrV <- Predict.matrix(model$scV,
data = data.frame(XPred,
Date2 = (time - mean(data$Date)) /
sd(data$Date)))
PredictionsSigma <-
rowMeans(sqrt(sapply(1:nrow(betaSigma), function(x)
exp((PredMatrV %*% betaSigma[x, ])) / model$model$sigma[x]) * model$sRe^2))
} else {
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
PredictionsSigma <- sqrt(Predictions * (1-Predictions))
} else {
PredictionsSigma <- sqrt(mean(model$model$sigma) * model$sRe^2)
}
}
if(estType == "Mean"){
Est <- rowMeans(Predictions)
}
if(estType == "1 SE"){
Est <- apply(Predictions, 1, sd)
}
if(estType == "2 SE"){
Est <- apply(Predictions, 1, sd) * 2
}
if(estType == "1 SETOTAL"){
Est <- sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2)
}
if(estType == "2 SETOTAL"){
Est <- sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2) * 2
}
if(estType == "1 SD Population"){
Est <- PredictionsSigma * 1
}
if(estType == "2 SD Population"){
Est <- PredictionsSigma * 2
}
if(estType == "Quantile"){
Est <- apply(Predictions, 1, quantile, estQuantile)
}
if(estType == "QuantileTOTAL"){
Est <- rowMeans(Predictions + qnorm(estQuantile) * sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2))
}
XPred <-
data.frame(XPred,
Est = Est,
Sd = apply(Predictions, 1, sd),
SDPop = PredictionsSigma,
SdTotal = sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2),
IntLower = apply(Predictions, 1, quantile, 0.025),
IntUpper = apply(Predictions, 1, quantile, 0.975),
IntLowerTotal = apply(Predictions + sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2), 1, quantile, 0.025),
IntUpperTotal = apply(Predictions + sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2), 1, quantile, 0.975),
resError = sqrt(mean(model$model$sigma + model$model$tau) * model$sRe^2))
}
if (Bayes == FALSE & GAM == FALSE){
Est <- predict(model$model$gam,
data.frame(XPred,
Date2 = (time - mean(data$Date)) /
sd(data$Date)), se.fit = TRUE, type = "response")
if(estType == "1 SE"){
Est$fit <- Est$se.fit
}
if(estType == "2 SE"){
Est$fit <- Est$se.fit * 2
}
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
varM = Est$fit * (1-Est$fit)
} else {
varM = var(residuals(model$model$gam))
}
if(estType == "1 SETOTAL"){
Est$fit <- sqrt(Est$se.fit^2 + varM)
}
if(estType == "2 SETOTAL"){
Est$fit <- sqrt(Est$se.fit^2 + varM) * 2
}
if(estType == "1 SD Population"){
Est$fit <- sqrt(varM) * 1
}
if(estType == "2 SD Population"){
Est$fit <- sqrt(varM) * 2
}
if(estType == "Quantile"){
Est$fit <- Est$fit + qnorm(estQuantile) * Est$se.fit
}
if(estType == "QuantileTOTAL"){
Est$fit <- Est$fit + qnorm(estQuantile) *
sqrt(Est$se.fit^2 + varM)
}
XPred <- data.frame(XPred,
Est = Est$fit, Sd = Est$se.fit,
SDPop = sqrt(varM),
SdTotal = sqrt(Est$se.fit^2 + varM),
IntLower = Est$fit - 1.96 * Est$se.fit,
IntUpper = Est$fit + 1.96 * Est$se.fit,
IntLowerTotal = Est$fit - 1.96 *
sqrt(Est$se.fit^2 + varM),
IntUpperTotal = Est$fit + 1.96 *
sqrt(Est$se.fit^2 + varM),
resError = sqrt(mean((predict(model$model$gam, type = "response") - model$model$gam$y)^2)))
}
if(GAM == TRUE){
Predictions <- sapply(1:length(model$model),
function(x) predict(model$model[[x]],
x = cbind(XPred$Longitude, XPred$Latitude,
Date2 = (time - mean(data$Date)) / sd(data$Date))))
if(estType == "Mean"){
Est <- rowMeans(Predictions)
}
if(estType == "1 SE"){
Est <- apply(Predictions, 1, sd)
}
if(estType == "2 SE"){
Est <- apply(Predictions, 1, sd) * 2
}
if(estType == "Quantile"){
Est <- apply(Predictions, 1, quantile, estQuantile)
}
XPred <- data.frame(XPred,
Est = Est,
Sd = apply(Predictions, 1, sd),
IntLower = apply(Predictions, 1, quantile, 0.025),
IntUpper = apply(Predictions, 1, quantile, 0.975))
}
if (estType != "1 SE" && estType != "1 SETOTAL" && estType != "2 SE" &&
estType != "2 SETOTAL" &&
!is.null(limitz) && !missing(limitz)){
if (limitz == "0-1"){
XPred$Est <- pmin(1, XPred$Est)
XPred$Est <- pmax(0, XPred$Est)
}
if (limitz == "0-100"){
XPred$Est <- pmin(100, XPred$Est)
XPred$Est <- pmax(0, XPred$Est)
}
}
if (!is.null(dataCenter)){
# this is batch mode
XPredCenter <- lapply(1:nrow(dataCenter), function(x){
XPred %>%
extractXPredCenter(centerX = dataCenter[x, 1],
centerY = dataCenter[x, 2],
Radius = RadiusBatch,
batch = TRUE,
isThreeD = TRUE,
data = data,
time = dataCenter[x, 3])
})
centerEstimates <- lapply(1:nrow(dataCenter), function(x){
XPredCenter[[x]] %>%
extractCenterEstimates(batch = FALSE) # why is sd calculated here differently than in plotMap batch??
})
meanCenter <- sapply(1:nrow(dataCenter), function(x) centerEstimates[[x]]$mean)
sdCenter <- sapply(1:nrow(dataCenter), function(x) centerEstimates[[x]]$sd)
IntLower <- sapply(1:nrow(dataCenter),
function(x) signif(min(XPredCenter[[x]]$IntLower), 5))
IntUpper <- sapply(1:nrow(dataCenter),
function(x) signif(max(XPredCenter[[x]]$IntUpper), 5))
if(is.null(XPredCenter[[1]]$SDPop)) {
dataCenter <- cbind(dataCenter, time = time, mean = meanCenter, sd = sdCenter,
IntLower = IntLower, IntUpper = IntUpper)
} else {
SDPop <- sapply(1:nrow(dataCenter),
function(x) signif(sqrt(mean(XPredCenter[[x]]$SDPop, na.rm = TRUE)^2), 5))
sdCenterTotal <- sapply(1:nrow(dataCenter),
function(x) signif(sqrt(sum(c(sd(XPredCenter[[x]]$Est)^2,
mean(XPredCenter[[x]]$Sd^2),
mean(XPredCenter[[x]]$SdTotal)^2), na.rm = TRUE)), 5))
IntLowerTotal <- sapply(1:nrow(dataCenter),
function(x) signif(min(XPredCenter[[x]]$IntLowerTotal), 5))
IntUpperTotal <- sapply(1:nrow(dataCenter),
function(x) signif(max(XPredCenter[[x]]$IntUpperTotal), 5))
dataCenter <- cbind(dataCenter, time = time, mean = meanCenter, sd = sdCenter,
SDPop = SDPop,
sdTotal = sdCenterTotal,
IntLower = IntLower, IntUpper = IntUpper,
IntLowerTotal = IntLowerTotal, IntUpperTotal = IntUpperTotal)
}
return(dataCenter)
}
# keep $Est for later calculation of mean and sd for center
XPred$EstForCenter <- XPred$Est
if (interior > 0){
# this can remove all $Est
XPred$Est[draw == 0] <- NA
}
if (mask == TRUE){
XPred$Est[maskDraw == 0] <- NA
}
if(GAM == TRUE){
XPred$Est[is.na(XPred$Est) | XPred$Est < 0] <- 0
}
#if (!all(is.na(XPred$Est))){
levels <- pretty(c(rangez[1], rangez[2]), n = ncol)
if(GAM == TRUE){
colors <- c("#FFFFFF", colorRampPalette(brewer.pal(9, colors))(length(levels) - 2))
} else{
colors <- colorRampPalette(brewer.pal(9, colors))(length(levels) - 1)
}
if(reverseColors){
colors <- rev(colors)
}
levelsLegend <- levels
if(length(levels) > 25){
par(fg = NA, col="black")
levelsLegend <- pretty(c(rangez[1], rangez[2]),
n = pmin(20, ceiling(ncol / 2)))
}
cex4 <- 1
if(!all(is.na(XPred$Est)) &&
(max(abs(XPred$Est), na.rm = TRUE) > 9999 | max(abs(XPred$Est), na.rm = TRUE) < 0.05)){
cex4 <- 0.7
}
if(centerMap != "Europe"){
longitudes <- longitudesPac[order(longitudesPac)]
latitudes <- latitudes[order(latitudes, longitudesPac)]
XPredPlot <- data.frame(XPredPac[order(XPredPac$Latitude, XPredPac$Longitude),],
Est = XPred$Est[order(XPredPac$Latitude, XPredPac$Longitude)])
rangex <- rangexEU
} else {
XPredPlot <- XPred
}
if(!showModel){
XPredPlot$Est <- rep(NA, length(XPredPlot$Est))
}
if(titleMain){
main = ""
} else {
if(setAxisLabels){
main = mainLabel
} else {
main = independent
}
}
if(titleScale){
mainS = ""
} else {
if(setAxisLabels){
mainS = scLabel
} else {
mainS = paste("time:", time)
}
}
if(setAxisLabels){
xlab = xLabel
ylab = yLabel
} else {
xlab = "Longitude"
ylab = "Latitude"
}
filled.contour2(longitudes, latitudes, z = matrix(XPredPlot$Est, ncol = resolution),
contourType = contourType,
xlim = rangex, ylim = rangey, levels = levels,
col = colors,
showScale = showScale,
cex.axis = 1.5, cex.main = 1.5, cex.lab = 1.5,
asp = 1, xaxp = c(0, 100, 10), yaxp = c(0, 100, 10),
key.axes = axis(side = 4, at = levelsLegend, cex.axis = cex4),
key.title = title(main = main, cex.main = 0.8),
plot.title = {title(cex.lab = AxisSize, xlab = xlab, ylab = ylab, main = mainS)},
plot.axes = {
par(fg = "black", col="black");
if (terrestrial == 1){
if(centerMap != "Europe"){
sp::plot(Maps$ocean160, add = T, col = "lightblue", lwd = 1, border = NA);
sp::plot(Maps$ocean200, add = T, col = "lightblue", lwd = 1, border = NA);
} else {
sp::plot(Maps$ocean, add = T, col = "lightblue", lwd = 1);
}
}
if (terrestrial == -1){
if(centerMap != "Europe"){
sp::plot(Maps$land160, add = T, lwd = 1, col = "grey96", border = NA);
sp::plot(Maps$land200, add = T, lwd = 1, col = "grey96", border = NA);
} else {
sp::plot(Maps$land, add = T, lwd = 1, col = "grey96", border = NA);
}
}
if(centerMap != "Europe"){
sp::plot(Maps$coast160, add = T, lwd = 1);
sp::plot(Maps$coast200, add = T, lwd = 1);
} else {
sp::plot(Maps$coast, add = T, lwd = 1);
}
if (grid == TRUE){
if(centerMap != "Europe"){
sp::plot(Maps$grids160, add = T, col = "grey", lty = 2, xlim = c(0, 1));
sp::plot(Maps$grids200, add = T, col = "grey", lty = 2, xlim = c(0, 1));
} else {
sp::plot(Maps$grids, add = T, col = "grey", lty = 2, xlim = c(0, 1));
}
}
if(centerMap != "Europe"){
sp::plot(Maps$borders160, add = T, col = "darkgrey", lwd = 1);
sp::plot(Maps$borders200, add = T, col = "darkgrey", lwd = 1);
} else {
sp::plot(Maps$borders, add = T, col = "darkgrey", lwd = 1);
}
# add points and/or centroids ----
if (points == TRUE){
if(!is.null(pointLabels)){
pointLabels <- as.numeric(pointLabels)
pointLabels <- pointLabels - min(pointLabels, na.rm = TRUE)
pointLabels <- pointLabels / max(pointLabels, na.rm = TRUE)
pointSize <- (pointSize * pointLabels + 0.1) * 2
}
if(!is.null(pointColLabels)){
pointColLabels <- as.numeric(pointColLabels)
pointColLabels <- pointColLabels - min(pointColLabels, na.rm = TRUE)
pointColLabels <- pointColLabels / max(pointColLabels, na.rm = TRUE)
pColor <- (colorRampPalette(brewer.pal(9, colorsP))(101))[round(pointColLabels, 2) * 100]
}
if(cluster & !is.null(data$spatial_cluster)){
data <- data %>%
selectClusterGrouping(cluster, clusterResults)
# get centroids
if(clusterResults == 1){
data_names <- c("spatial_cluster", "long_centroid_spatial_cluster", "lat_centroid_spatial_cluster", "cluster")
centroids <- unique(data[, data_names])
centroids <- na.omit(centroids)
} else {
data_names <- c("temporal_group", "long_temporal_group_reference_point", "lat_temporal_group_reference_point", "cluster")
centroids <- unique(data[, data_names])
}
centroids <- centroids[order(centroids[,1]), ]
if(centerMap != "Europe"){
centroidsPac <- centroids
centroidsPac[,2][centroids[,2] < -20] <- centroidsPac[,2][centroids[,2] < -20] + 200
centroidsPac[,2][centroids[,2] >= -20] <- (- 160 + centroidsPac[,2][centroids[,2] >= -20])
centroids <- centroidsPac
}
}
# add points (location marks) ----
## if there is no clustering, or
## if not show only centroids
if (!cluster || clusterAll != "-1") {
dataPlot <- data %>%
centerPlotData(centerMap = centerMap)
## if there is no clustering, or
## if not "Show points for all times"
if (!cluster || clusterAll == "1") {
dataPlot <- dataPlot %>%
filterT(addU = addU, time = time)
}
dataPlot <- dataPlot %>%
selectClusterGrouping(cluster, clusterResults)
points(dataPlot$Latitude ~ dataPlot$Longitude,
col = getPColor(dataPlot, cluster, palName = clusterCol, pColor = pColor),
lwd = 2,
pch = pointShape,
cex = pointSize)
}
# add centroids ----
if (cluster & !is.null(data$spatial_cluster)) {
points(centroids[, 2:3],
col = getPColor(centroids, cluster, palName = clusterCol, pColor = pColor),
lwd = 2,
pch = pointShape,
cex = pointSize * 2.5)
text(centroids[, 2:3],
labels = sprintf("%s_%s",
ifelse(clusterResults == "0", "Group", "Cluster"),
centroids$cluster),
pos = 4,
cex = fontSize * 1.5,
col = fontCol,
family = fontType)
}
}
# add custom labels ----
if (!is.null(textLabels)) {
if (centerMap != "Europe") {
text(dataPac$Latitude ~ dataPac$Longitude,
labels = as.character(textLabels), pos = 4,
cex = fontSize, col = fontCol, family = fontType)
} else {
text(data$Latitude ~ data$Longitude,
labels = as.character(textLabels), pos = 4,
cex = fontSize, col = fontCol, family = fontType)
}
}
# add custom points ----
if (!is.null(pointDat) & NROW(pointDat) > 0) {
points(x = pointDat$x, y = pointDat$y, cex = pointDat$pointSize, col = pointDat$pointColor, pch = 16)
text(pointDat$y ~ pointDat$x, labels = pointDat$label, pos = 4, cex = 1.75)
}
# update axis if center Pacific
if (centerMap != "Europe") {
lab <- pretty(rangex)
lab[pretty(rangex) >= 20] <- lab[pretty(rangex) >= 20] - 200
lab[pretty(rangex) < 20] <- lab[pretty(rangex) < 20] + 160
axis(1, at = pretty(rangex), labels = lab, cex.axis = AxisLSize);
} else{
axis(1, cex.axis = AxisLSize);
}
axis(2, cex.axis = AxisLSize);
}
)
# } else {
# plot(1:10, xlim = range(data$Longitude), ylim = range(data$Latitude),
# xlab = "Longitude", ylab = "Latitude")
# plot(Maps$ocean, add = T, col = "lightblue", lwd = 1)
# sp::plot(Maps$grids, add = T, col = "grey", lty = 2, xlim = c(0, 1))
# sp::plot(Maps$borders, add = T, col = "darkgrey", lwd = 1)
# text(mean(data$Longitude), mean(data$Latitude), cex = 1,
# "No estimates to plot, please adjust
# \"Display up to max standard error\" slider", col = "red")
# }
if (arrow == TRUE){
if(showScale == TRUE){
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
} else {
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
}
}
if (scale == TRUE){
if(showScale == TRUE){
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
} else {
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
}
}
if(plotRetNull){
return(NULL)
} else {
return(list(XPred = XPred))
}
}
#' Plots difference or similarity map
#'
#' @param XPred return object of similarityMap or createDifferenceMap functions
#' @param estType one of "Mean", "SE" and "Quantile". defaults to "Mean"
#' @param estQuantile quantile (only applicable if estType = "Quantile")
#' @param type one of "similarity" or "difference" determining the type of map
#' @param independent name of independent variable shown in plot
#' @param arrow display north arrow TRUE/FALSE
#' @param scale display scale TRUE/FALSE
#' @param terrestrial show only estimates on land masses (1), oceans (-1) or all (0)
#' @param grid show coordinate grid TRUE/FALSE
#' @param ncol number of colors for estimates
#' @param colors color scheme of estimates from RColorBrewer. defaults to "RdYlGn"
#' @param reverseColors inverse colour scheme
#' @param rangex range of longitude values (x axis limits)
#' @param rangey range of latitude values (y axis limits)
#' @param rangez range of estimated values (z axis limits)
#' @param centerMap center of map, one of "Europe" and "Pacific"
#' @param showValues boolean show values in plot?
#' @param simValues if showValues: list of simulated values
#' @param dataCenter data for batch estimation
#' @param RadiusBatch radius
#' @param showModel show model
#' @param titleMain show main title
#' @param titleScale show scale title
#' @param setAxisLabels show axis/main/scale labels
#' @param mainLabel main label
#' @param yLabel y lab label
#' @param xLabel y lab label
#' @param scLabel scale lab labels
#' @param northSize size of north arrow
#' @param scalSize size of scale
#' @param scaleX scale x orientation
#' @param scaleY scale y orientation
#' @param NorthX north arrow x orientation
#' @param NorthY north arrow y orientation
#' @param AxisSize axis title font size
#' @param AxisLSize axis label font size
#' @param pointDat data frame of points to add to plot
#' @inheritParams filled.contour2
#'
#' @export
plotDS <- function(XPred,
estType = "mean",
estQuantile = 0.95,
type = "similarity",
independent = "",
rangex = range(XPred$Longitude, na.rm = TRUE),
rangey = range(XPred$Latitude, na.rm = TRUE),
rangez = range(XPred$Est, na.rm = TRUE),
contourType = c("filled.contour", "contour"),
showModel = TRUE,
showScale = TRUE,
ncol = 10, colors = "RdYlGn",
centerMap = "Europe",
reverseColors = FALSE, terrestrial = 1, grid = TRUE,
arrow = TRUE, scale = TRUE,
simValues = NULL,
showValues = FALSE,
dataCenter = NULL, RadiusBatch = NULL,
titleMain = TRUE,
titleScale = TRUE,
setAxisLabels = FALSE,
mainLabel = "",
yLabel = "Latitude",
xLabel = "Longitude",
scLabel = "",
northSize = 0.2,
scalSize = 0.1,
scaleX = 0,
scaleY = 0.1,
NorthX = 0.025,
NorthY = 0.925,
AxisSize = 1,
AxisLSize = 1,
pointDat = NULL){
options(scipen = 999)
contourType <- match.arg(contourType)
RadiusBatch <- RadiusBatch / 111
minRangeFactor <- 0.75
if ((diff(rangex) / diff(rangey)) < minRangeFactor) {
rangex[1] <- max(-180, mean(rangex) - minRangeFactor / 2 * diff(rangey))
rangex[2] <- min(180, mean(rangex) + minRangeFactor / 2 * diff(rangey))
}
minRangeFactor <- 0.5
if((diff(rangey) / diff(rangex)) < minRangeFactor){
rangey[1] <- max(-90, mean(rangey) - minRangeFactor / 2 * diff(rangex))
rangey[2] <- min(90, mean(rangey) + minRangeFactor / 2 * diff(rangex))
}
if (!is.null(dataCenter)){
# this is batch mode
XPred2 <- do.call("rbind", lapply(1:nrow(dataCenter), function(x) {
longitudes <- seq(dataCenter[x, 1] - RadiusBatch,
dataCenter[x, 1] + RadiusBatch, length.out = 4)
latitudes <- seq(dataCenter[x, 2] - RadiusBatch,
dataCenter[x, 2] + RadiusBatch, length.out = 4)
XPred2 <- expand.grid(Longitude = longitudes, Latitude = latitudes)
cbind(XPred2, id = x)
}))
closestMean <- sapply(1:nrow(XPred2), function(x)
XPred$Est[which.min((XPred2$Longitude[x] - XPred$Longitude) ^ 2 +
(XPred2$Latitude[x] - XPred$Latitude) ^ 2
)])
closestSD <- sapply(1:nrow(XPred2), function(x)
XPred$Sd[which.min((XPred2$Longitude[x] - XPred$Longitude) ^ 2 +
(XPred2$Latitude[x] - XPred$Latitude) ^ 2
)])
closestDist <- sapply(1:nrow(XPred2), function(x)
min((XPred2$Longitude[x] - XPred$Longitude) ^ 2 +
(XPred2$Latitude[x] - XPred$Latitude) ^ 2
))
closestMean[closestDist > RadiusBatch / 5] <- NA
closestSD[closestDist > RadiusBatch / 5] <- NA
XPred2$Mean <- closestMean
XPred2$SD <- closestSD
XPred2 <- do.call("rbind", lapply(1:max(XPred2$id), function(x) {
tmp <- XPred2[XPred2$id == x,]
data.frame(
Longitude = tmp$Longitude[1],
Latitude = tmp$Latitude[1],
mean = signif(mean(tmp$Mean, na.rm = TRUE), 3),
sd = signif(mean(tmp$SD, na.rm = TRUE) +
sd(tmp$Mean, na.rm = TRUE), 3)
)
})
)
XPred2$IntLower <- XPred2$mean - 1.96 * XPred2$sd
XPred2$IntUpper <- XPred2$mean + 1.96 * XPred2$sd
return(XPred2)
}
if(estType == "1 SE"){
XPred$Est <- XPred$Sd
}
if(estType == "2 SE"){
XPred$Est <- 2 * XPred$Sd
}
if(estType == "Significance (p-value)"){
XPred$Est <- pnorm(-abs(XPred$Est), 0, XPred$Sd) * 2
}
if(estType == "Significance (z-value)"){
XPred$Est <- XPred$Est / XPred$Sd
}
if(estType == "Quantile"){
XPred$Est <- XPred$Est + qnorm(estQuantile) * XPred$Sd
if(type == "similarity"){
XPred$Est <- pmax(0, XPred$Est)
}
}
#if (!all(is.na(XPred$Est))){
if(type == "similarity"){
z <- matrix(XPred$Est, ncol = length(unique(XPred$Latitude)))
}
if(type == "difference"){
z <- matrix(XPred$Est, ncol = length(unique(XPred$Latitude)))
}
Maps <- loadMaps()
levels <- pretty(c(rangez[1], rangez[2]), n = ncol)
colors <- colorRampPalette(brewer.pal(9, colors))(length(levels) - 1)
if(reverseColors){
colors <- rev(colors)
}
levelsLegend <- levels
if(length(levels) > 25){
par(fg = NA, col="black")
levelsLegend <- pretty(c(rangez[1], rangez[2]),
n = pmin(20, ceiling(ncol / 2)))
}
# keep $Est for later calculation of mean and sd for center
XPred$EstForCenter <- XPred$Est
if(centerMap != "Europe"){
XPredPac <- XPred
XPredPac$Longitude[XPred$Longitude < -20] <- XPredPac$Longitude[XPred$Longitude < -20] + 200
XPredPac$Longitude[XPred$Longitude >= -20] <- (- 160 + XPredPac$Longitude[XPred$Longitude >= -20])
XPredPlot <- data.frame(XPredPac[order(XPredPac$Latitude, XPredPac$Longitude),])
z <- matrix(XPredPlot$Est, ncol = length(unique(XPredPlot$Latitude)))
} else {
XPredPlot <- XPred
}
if(!showModel){
z <- matrix(NA, ncol = length(unique(XPredPlot$Latitude)), nrow = length(unique(XPredPlot$Longitude)))
}
cex4 <- 1
if(max(abs(z), na.rm = TRUE) > 9999 | max(abs(z), na.rm = TRUE) < 0.05){
cex4 <- 0.7
}
if(titleMain){
main = ""
} else {
if(setAxisLabels){
main = mainLabel
} else {
main = independent
}
}
if(titleScale){
mainS = ""
} else {
if(setAxisLabels){
mainS = scLabel
} else {
mainS = independent
}
}
if(setAxisLabels){
xlab = xLabel
ylab = yLabel
} else {
xlab = "Longitude"
ylab = "Latitude"
}
if (any(unique(XPredPlot$Longitude) != sort(unique(XPredPlot$Longitude))) |
any(unique(XPredPlot$Latitude) != sort(unique(XPredPlot$Latitude)))){
return("The meridian should match the corresponding region. Spherical coordinates should also match. Try to adjust the map centering option.")
}
filled.contour2(unique(XPredPlot$Longitude), unique(XPredPlot$Latitude),
z = z,
contourType = contourType,
xlim = rangex, ylim = rangey, levels = levels,
col = colors,
showScale = showScale,
cex.axis = 1.5, cex.main = 1.5, cex.lab = 1.5,
asp = 1, key.axes = axis(side = 4, at = levelsLegend, cex.axis = cex4),
key.title = title(main = main, cex.main = 0.8),
plot.title = {title(cex.lab = AxisSize, xlab = xlab, ylab = ylab, main = mainS)},
plot.axes = {
par(fg = "black", col="black");
if (terrestrial == 1){
if(centerMap != "Europe"){
sp::plot(Maps$ocean160, add = T, col = "lightblue", lwd = 1, border = NA);
sp::plot(Maps$ocean200, add = T, col = "lightblue", lwd = 1, border = NA);
} else {
sp::plot(Maps$ocean, add = T, col = "lightblue", lwd = 1);
}
}
if (terrestrial == -1){
if(centerMap != "Europe"){
sp::plot(Maps$land160, add = T, lwd = 1, col = "grey96", border = NA);
sp::plot(Maps$land200, add = T, lwd = 1, col = "grey96", border = NA);
} else {
sp::plot(Maps$land, add = T, lwd = 1, col = "grey96", border = NA);
}
}
if(centerMap != "Europe"){
sp::plot(Maps$coast160, add = T, lwd = 1);
sp::plot(Maps$coast200, add = T, lwd = 1);
} else {
sp::plot(Maps$coast, add = T, lwd = 1);
}
if (grid == TRUE){
if(centerMap != "Europe"){
sp::plot(Maps$grids160, add = T, col = "grey", lty = 2, xlim = c(0, 1));
sp::plot(Maps$grids200, add = T, col = "grey", lty = 2, xlim = c(0, 1));
} else {
sp::plot(Maps$grids, add = T, col = "grey", lty = 2, xlim = c(0, 1));
}
}
if(centerMap != "Europe"){
sp::plot(Maps$borders160, add = T, col = "darkgrey", lwd = 1);
sp::plot(Maps$borders200, add = T, col = "darkgrey", lwd = 1);
} else {
sp::plot(Maps$borders, add = T, col = "darkgrey", lwd = 1);
}
if(centerMap != "Europe"){
lab <- pretty(rangex)
lab[pretty(rangex) >= 20] <- lab[pretty(rangex) >= 20] - 200
lab[pretty(rangex) < 20] <- lab[pretty(rangex) < 20] + 160
axis(1, at = pretty(rangex), labels = lab, cex.axis = AxisLSize);
} else{
axis(1, cex.axis = AxisLSize);
}
axis(2, cex.axis = AxisLSize);
if(!is.null(pointDat) & NROW(pointDat) > 0){
points(x = pointDat$x, y = pointDat$y, cex = pointDat$pointSize, col = pointDat$pointColor, pch = 16)
text(pointDat$y ~ pointDat$x, labels = pointDat$label, pos = 4, cex = 1.75)
}
if (showValues == TRUE) {
op <- par(family = "mono")
titel = paste(capture.output(do.call("rbind", simValues)), collapse = '\n')
legend("bottomright", legend = c(""), pch = NA, title = titel,
text.width = strwidth(titel)[1]/2, bty = "n", cex = 2, text.font = 2)
par(op)
}
}
)
# } else {
# plot(1:10, xlim = range(XPred$Longitude), ylim = range(XPred$Latitude),
# xlab = "Longitude", ylab = "Latitude")
# plot(Maps$ocean, add = T, col = "lightblue", lwd = 1)
# sp::plot(Maps$grids, add = T, col = "grey", lty = 2, xlim = c(0, 1))
# sp::plot(Maps$borders, add = T, col = "darkgrey", lwd = 1)
# text(mean(XPred$Longitude), mean(XPred$Latitude), cex = 1,
# "No estimates to plot, please adjust \"Display up to max standard error\" slider",
# col = "red")
# }
if (arrow == TRUE){
if(showScale == TRUE){
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
} else {
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
}
}
if (scale == TRUE){
if(showScale == TRUE){
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
} else {
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
}
}
return(list(XPred = XPred))
}
#' Filled Countour 2
#' Wrapper for contour plot
#'
#' @param x x values
#' @param y y values
#' @param z z values
#' @param contourType one of "filled.contour" or "contour"
#' @param xlim x limits
#' @param ylim y limits
#' @param zlim z limits
#' @param levels levels
#' @param nlevels number of levels
#' @param showScale show colour scale
#' @param color.palette color palette
#' @param col colors
#' @param plot.title plot title
#' @param plot.axes plot axes
#' @param key.title key title
#' @param key.axes key axes
#' @param asp aspect ratio
#' @param xaxs x axis style
#' @param yaxs y axis style
#' @param las label style
#' @param axes show axes
#' @param frame.plot frame plot
#' @param ... additional arguments
filled.contour2 <- function(x = seq(0, 1, length.out = nrow(z)),
y = seq(0, 1, length.out = ncol(z)),
z,
contourType = c("filled.contour", "contour"),
xlim = range(x, finite = TRUE),
ylim = range(y, finite = TRUE), zlim = range(z, finite = TRUE),
levels = pretty(zlim, nlevels), nlevels = 20, showScale = TRUE,
color.palette = cm.colors, col = color.palette(length(levels) - 1),
plot.title, plot.axes, key.title, key.axes, asp = NA, xaxs = "i",
yaxs = "i", las = 1, axes = TRUE, frame.plot = axes, ...)
{
contourType <- match.arg(contourType)
if (missing(z)) {
if (!missing(x)) {
if (is.list(x)) {
z <- x$z
y <- x$y
x <- x$x
}
else {
z <- x
x <- seq.int(0, 1, length.out = nrow(z))
}
}
else stop("no 'z' matrix specified")
}
else if (is.list(x)) {
y <- x$y
x <- x$x
}
if (any(diff(x) <= 0) || any(diff(y) <= 0))
stop("increasing 'x' and 'y' values expected")
# setup layout
mar.orig <- (par.orig <- par(c("mar", "las", "mfrow")))$mar
on.exit(par(par.orig))
par(las = las)
pin1 <- par("pin")
ratioLim <- abs(diff(ylim)) / abs(diff(xlim))
if (showScale && contourType == "filled.contour") {
# set up colour legend
w <- (3 + mar.orig[2L]) * par("csi") * 2.54
layout(matrix(c(2, 1), ncol = 2L), widths = c(1, lcm(w)))
mar <- mar.orig
mar[4L] <- mar[2L]
mar[2L] <- 1
par(mar = mar)
a = (pin1[1] + par("mai")[2] + par("mai")[4])
b = (pin1[2] + par("mai")[1] + par("mai")[3])
ratioXY <- (a / b)
if ((1 / ratioLim) >= ratioXY) {
par(plt = getPlotRegion(bottom = 0.5, ratioLim = ratioLim, ratioXY = ratioXY))
}
if ((1 / ratioLim) < ratioXY) {
par(plt = c(0.15, 0.5, 0.15, 0.9))
}
# create colour legend
plot.new()
plot.window(xlim = c(0, 1), ylim = range(levels), xaxs = "i",
yaxs = "i")
rect(0, levels[-length(levels)], 1, levels[-1L], col = col)
if (missing(key.axes)) {
if (axes)
if (max(abs(z), na.rm = TRUE) < 10000) {
axis(4)
} else {
axis(4, cex.axis = 0.7)
}
}
else key.axes
box()
if (!missing(key.title))
key.title
}
# set up (filled.)contour
mar <- mar.orig
mar[4L] <- 1
par(mar = mar)
a = (pin1[1] + par("mai")[2] + par("mai")[4])
b = (pin1[2] + par("mai")[1] + par("mai")[3])
ratioXY <- (a / b)
if ((1 / ratioLim) >= ratioXY) {
par(plt = getPlotRegion(bottom = 0.975, ratioLim = ratioLim, ratioXY = ratioXY))
}
if ((1 / ratioLim) < ratioXY) {
add <- 1 / ratioXY / 2 * 0.75 / ratioLim
par(plt = c(0.975 - 2 * add, 0.975, 0.15, 0.9))
}
# create (filled.)contour
plot.new()
if (contourType == "contour") {
# contour plot ----
contour(x = x, y = y, z = z, nlevels = length(levels), col = col,
xlim = xlim, ylim = ylim, xaxs = xaxs, yaxs = yaxs, asp = asp)
} else {
# filled.contour plot ----
plot.window(xlim, ylim, "", xaxs = xaxs, yaxs = yaxs, asp = asp)
.filled.contour(x, y, z, levels, col)
}
# set plot axis and background
if (missing(plot.axes)) {
if (axes) {
title(main = "", xlab = "", ylab = "")
Axis(x, side = 1)
Axis(y, side = 2)
}
}
else plot.axes
if (frame.plot)
box()
if (missing(plot.title))
title(...)
else plot.title
invisible()
}
getPlotRegion <- function(left = 0.1, bottom = 0.975, ratioLim = 1, ratioXY = 1) {
c(left, bottom, 0.525 - ratioLim * ratioXY / 2 * 0.875, 0.525 + ratioLim * ratioXY / 2 * 0.875)
}
north.arrow = function(x, y, h, c, adj) {
polygon(c(x, x, x + h/2), c(y - h, y, y - (1 + sqrt(3)/2) * h),
col = "black", border = NA)
polygon(c(x, x + h/2, x, x - h/2), c(y - h, y - (1 + sqrt(3)/2) *
h, y, y - (1 + sqrt(3)/2) * h))
text(x, y, "N", adj = adj, cex = c)
}
#' Plots time course map of a spatio-temporal model from estimateMap3D() function
#'
#' @param model return object of a spatio-temporal model from estimateMap3D() function
#' @param IndSelect for categorical model: selected category
#' @param independent name of independent variable shown in plot
#' @param trange range of longitude values (x axis limits)
#' @param resolution temporal grid resolution of displayed (higher is slower but better quality)
#' @param centerX longitude value to display time course plot for
#' @param centerY latitude value to display time course plot for
#' @param rangey 2-element vector of time interval to show
#' @param seType setype
#' @param pointDat add points/lines to plot
#' @param pointsTime should nearby points be plotted?
#' @param returnPred should the prediction be returned?
#' @param intTime should uncertainty intervals of nearby points be plotted?
#' @param rangePointsTime range of nearby points in km
#' @param limitz z limit range
#' @param formatTimeCourse parameters for the plot format, e.g. axesDecPlace, nLabelsX, nLabelsY
#'
#' @export
plotTimeCourse <- function(model, IndSelect = NULL,
independent = "", trange = range(model$data$Date),
resolution = 500, centerX = NA,
centerY = NA, rangey = NULL,
seType = "2",
pointDat = NULL,
pointsTime = FALSE,
returnPred = FALSE,
intTime = FALSE,
rangePointsTime = 500,
limitz = NULL,
formatTimeCourse = NULL){
sdValue <- 1
if(as.numeric(seType) > 5){
sdValue <- 2
}
minVal <- -Inf
maxVal <- Inf
if(!is.null(limitz) && limitz == "0-100"){
minVal <- 0
maxVal <- 100
}
if(!is.null(limitz) && limitz == "0-1"){
minVal <- 0
maxVal <- 1
}
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
if(IndSelect == "" | is.null(IndSelect)){
return(NULL)
}
model$model <- model$model[[IndSelect]]
minVal <- max(minVal, 0)
maxVal <- min(maxVal, 1)
}
Bayes = TRUE
GAM = FALSE
if("gamm" %in% class(model$model)){
Bayes = FALSE
}
if("kde" %in% class(model$model)){
GAM = TRUE
}
data <- model$data
if (is.null(data)) return(NULL)
if (is.na(centerX)){centerX <- signif(mean(data$Longitude, na.rm = TRUE), 3)}
if (is.na(centerY)){centerY <- signif(mean(data$Latitude, na.rm = TRUE), 3)}
time <- seq(trange[1], trange[2], length.out = resolution)
XPred <- data.frame(time,
Date2 = (time - mean(data$Date)) /
sd(data$Date),
Longitude2 = (centerX - mean(data$Longitude)) / sd(data$Longitude),
Latitude2 = (centerY - mean(data$Latitude)) / sd(data$Latitude),
Longitude = centerX,
Latitude = centerY)
if (Bayes == TRUE & GAM == FALSE){
sc <- model$sc
PredMatr <- Predict.matrix(sc, data = XPred)
betas <- model$model$beta
betaSigma <- model$model$betaSigma
Predictions <-
sapply(1:nrow(betas), function(x)
PredMatr %*% betas[x, ] * model$sRe + model$mRe)
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
Predictions <- invLogit(Predictions)
}
if(!is.null(betaSigma)){
PredMatrV <- Predict.matrix(model$scV, data = XPred)
PredictionsSigma <-
rowMeans(sqrt(sapply(1:nrow(betaSigma), function(x)
exp((PredMatrV %*% betaSigma[x, ])) / model$model$sigma[x]) * model$sRe^2))
} else {
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
PredictionsSigma <- sqrt(Predictions * (1-Predictions))
} else {
PredictionsSigma <- sqrt(mean(model$model$sigma) * model$sRe^2)
}
}
if(!is.null(betaSigma)){
SdTotal <- sqrt((PredictionsSigma)^2 + apply(Predictions, 1, sd)^2)
} else {
SdTotal <- sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2)
}
XPred <-
data.frame(XPred,
Est = rowMeans(Predictions),
Sd = apply(Predictions, 1, sd),
SdTotal = SdTotal,
PredictionsSigma = PredictionsSigma,
IntLower = pmax(minVal, pmin(maxVal, rowMeans(Predictions) - sdValue * apply(Predictions, 1, sd))),
IntUpper = pmax(minVal, pmin(maxVal, rowMeans(Predictions) + sdValue * apply(Predictions, 1, sd))))
mainlab <- paste0("Estimate of ", independent, " in time course at coordinates ",
"(", centerY,",", centerX,")" ," with credible intervals")
}
if (Bayes == FALSE & GAM == FALSE){
Est <- predict(model$model$gam, XPred, se.fit = TRUE, type = "response")
XPred <- data.frame(XPred, Est = Est$fit, Sd = Est$se.fit,
SdTotal = sqrt(Est$se.fit^2 + mean(residuals(model$model$gam)^2)),
PredictionsSigma = sd(residuals(model$model$gam)),
IntLower = pmax(minVal, pmin(maxVal, Est$fit - sdValue * Est$se.fit)),
IntUpper = pmax(minVal, pmin(maxVal, Est$fit + sdValue * Est$se.fit))
# IntLowerTotal = Est$fit - 1.96 * sqrt(Est$se.fit^2 + mean(residuals(model$model$gam)^2)),
# IntUpperTotal = Est$fit + 1.96 * sqrt(Est$se.fit^2 + mean(residuals(model$model$gam)^2))
)
mainlab <- paste0("Estimate of ", independent, " in time course at coordinates ",
"(", centerY,",", centerX,")" ," with confidence intervals")
}
if (GAM == TRUE){
EstTemp <- sapply(1:length(model$model), function(x) predict(model$model[[x]],
x = cbind(XPred$Longitude, XPred$Latitude, XPred$Date2)))
estQuantile <-
XPred <- data.frame(XPred,
Est = rowMeans(EstTemp),
Sd = apply(EstTemp, 1, sd),
IntLower = pmax(0, apply(EstTemp, 1, quantile, 1 - pnorm(sdValue))),
IntUpper = pmax(0, apply(EstTemp, 1, quantile, pnorm(sdValue))))
mainlab <- paste0("Density estimate in time course at coordinates ",
"(", centerY,",", centerX,").")
}
ylims <- c(min(XPred$Est - sdValue * XPred$Sd), max(XPred$Est + sdValue * XPred$Sd))
if(!is.null(rangey)){
ylims[!is.na(rangey)] = rangey[!is.na(rangey)]
}
plot(
pmax(minVal, pmin(maxVal, XPred$Est)) ~ time,
type = "l",
ylim = ylims,
xlim = c(min(time), max(time)),
ylab = independent,
xlab = "time",
main = mainlab,
xaxt = 'n',
yaxt = 'n'
)
if (!is.null(formatTimeCourse)) {
addFormattedAxis(
axis = "x",
min = min(time),
max = max(time),
nLabels = formatTimeCourse$nLabelsX,
decPlace = formatTimeCourse$axesDecPlace
)
addFormattedAxis(
axis = "y",
min = ylims[1],
max = ylims[2],
nLabels = formatTimeCourse$nLabelsY,
decPlace = formatTimeCourse$axesDecPlace
)
}
if(seType %in% c("2", "4", "6", "9")){
polygon(c(rev(time), time), pmax(minVal, pmin(maxVal, c(rev(XPred$IntUpper), XPred$IntLower))),
col = 'grey90', border = NA)
lines(pmax(minVal, pmin(maxVal, (XPred$Est))) ~ (time), lty = 1)
lines(pmax(minVal, pmin(maxVal, (XPred$IntUpper))) ~ (time), lty = 2)
lines(pmax(minVal, pmin(maxVal, (XPred$IntLower))) ~ (time), lty = 2)
}
if(seType %in% c("5", "7")){
polygon(c(rev(time), time), pmax(minVal, pmin(maxVal,
c(rev(XPred$Est - sdValue * XPred$PredictionsSigma),
XPred$Est + sdValue * XPred$PredictionsSigma))),
col = 'grey90', border = NA)
lines(pmax(minVal, pmin(maxVal, (XPred$Est))) ~ (time), lty = 1)
lines(pmax(minVal, pmin(maxVal, (XPred$Est - sdValue * XPred$PredictionsSigma))) ~ (time), lty = 2)
lines(pmax(minVal, pmin(maxVal, (XPred$Est + sdValue * XPred$PredictionsSigma))) ~ (time), lty = 2)
}
if(seType %in% c("3", "4", "8", "9")){
lines(pmax(minVal, pmin(maxVal,(XPred$Est + sdValue * XPred$SdTotal))) ~ (time), lty = 3)
lines(pmax(minVal, pmin(maxVal,(XPred$Est - sdValue * XPred$SdTotal))) ~ (time), lty = 3)
}
if(!is.null(pointDat) & NROW(pointDat) > 0){
for(i in 1:nrow(pointDat)){
if(!is.na(pointDat$y[i]) & !is.na(pointDat$x[i])){
points(pointDat$y[i] ~ pointDat$x[i], col = pointDat$pointColor[i],
cex = pointDat$pointSize[i], pch = 19)
if(!is.na(pointDat$ymin[i]) & !is.na(pointDat$ymax[i])){
lines(x = rep(pointDat$x[i], 2), y = c(pointDat$ymin[i], pointDat$ymax[i]),
lwd = pointDat$pointSize[i])
}
if(!is.na(pointDat$xmin[i]) & !is.na(pointDat$xmax[i])){
lines(y = rep(pointDat$y[i], 2), x = c(pointDat$xmin[i], pointDat$xmax[i]),
lwd = pointDat$pointSize[i])
}
}
}
}
if(returnPred){
return(XPred[, !(names(XPred) %in% c("Date2", "Longitude2", "Latitude2"))])
}
if(pointsTime){
pointPlotData <- model$data[(sqrt((model$data$Longitude - centerX)^2 +
(model$data$Latitude - centerY)^2) * 111) < rangePointsTime, ]
if(NROW(pointPlotData) > 0){
if(model$independent == ""){
pointPlotData$ind <- 0
ind <- "ind"
} else {
if(!is.null(model$IndependentType) && model$IndependentType == "numeric"){
ind <- model$independent
} else {
if(IndSelect == "" | is.null(IndSelect)){
ind <- model$independent
} else {
ind <- IndSelect
}
}
}
points(pointPlotData[, ind] ~ pointPlotData$Date)
if(intTime){
for(i in 1:nrow(pointPlotData)){
lines(c((pointPlotData$Date[i] - 2 * pointPlotData$Uncertainty[i]),
(pointPlotData$Date[i] + 2 * pointPlotData$Uncertainty[i])),
c(pointPlotData[, ind][i], pointPlotData[, ind][i]))
}
pointPlotData <- pointPlotData[(pointPlotData$Date + 2 * pointPlotData$Uncertainty > trange[1]) &
(pointPlotData$Date - 2 * pointPlotData$Uncertainty < trange[2]), ]
} else {
pointPlotData <- pointPlotData[(pointPlotData$Date > trange[1]) &
(pointPlotData$Date < trange[2]), ]
}
}
return(pointPlotData)
}
return(NULL)
}
#' Add Formatted Axis
#'
#' @param axis (character) axis to add, either "x" or "y"
#' @param min (numeric) position of 1st label
#' @param max (numeric) position of last label
#' @param nLabels (numeric) number of displayed labels
#' @param decPlace (numeric) number of the label's decimal places
addFormattedAxis <- function(axis, min, max, nLabels = 7, decPlace = 0) {
labelPositions <- seq(min, max, length.out = nLabels)
axisType <- switch(axis,
"x" = 1,
"y" = 2)
axis(axisType,
at = labelPositions,
labels = sprintf(paste('%1.', decPlace, 'f', sep = ""), labelPositions))
return(NULL)
}
plotTimeIntervals <- function(Model,
trange = c(0, 1000),
AxisSize = 1,
AxisLSize = 1,
cluster = FALSE,
clusterCol = "Set1",
clusterResults = 0
){
dat <- Model$data
dat <- dat %>%
selectClusterGrouping(cluster = cluster,
clusterResults = clusterResults)
if(!is.null(dat$spatial_cluster)){
dat$cluster_color <- colorRampPalette(brewer.pal(8, clusterCol))(max(dat$cluster, na.rm=TRUE))[dat$cluster]
dat <- dat[!is.na(dat$cluster),]
dat$cluster <- factor(dat$cluster)
g <- ggplot(dat, aes_(~Date, ~cluster)) + theme_light() + coord_cartesian(xlim = trange) +
theme(panel.grid.major.x = element_blank(),
panel.grid.minor.x = element_blank(),
axis.text=element_text(size=12 * AxisLSize),
axis.title=element_text(size=14 * AxisSize), legend.position = "none") +
geom_point(color=dat$cluster_color, position = position_dodge(0.3), alpha = 0.3) +
geom_errorbar(
aes_(xmin = ~ Date-2*Uncertainty, xmax = ~ Date + 2*Uncertainty),
color=dat$cluster_color,
position = position_dodge(0.3), width = 0.1, alpha = 0.3) +
ylab(ifelse(clusterResults == "0", "temporal_group", "spatial_cluster"))
print(g)
} else {
plot(1, cex = 0.1)
text(x = 1, y= 1, "Please run KernelTimeR with Cluster Analysis (left Panel)")
}
}
roundUpNice <- function(x, nice=c(1,2,5,10)) {
y <- x
if(y < 0){
x <- -x
}
if(length(x) != 1) stop("'x' must be of length 1")
x <- 10^floor(log10(x)) * nice[[which(x <= 10^floor(log10(x)) * nice)[[1]]]]
if(y < 0){
x <- -x
}
x
}
createSimilarityMap <- function(XPredList, pointList,
includeUncertainty = TRUE,
normalize = FALSE,
normalType = "1",
weightProb = FALSE,
weightMap = NULL,
negZero = TRUE,
invWeight = FALSE,
weightDecay = 1){
XPredList <- lapply(1:length(XPredList), function(x){
similarityMap(XPredList[[x]], pointList[[x]], includeUncertainty)
})
combineSimilarityMaps(XPredList,
normalize = normalize,
normalType = normalType,
weightProb = weightProb,
weightMap = weightMap,
negZero = negZero,
invWeight = invWeight,
weightDecay = weightDecay)
}
similarityMap <- function(XPred, point, includeUncertainty = TRUE) {
if (includeUncertainty == TRUE) {
density <- lapply(1:nrow(point), function(x)
log(rowMeans(sapply(1:1000,
function(y)
dnorm(
rnorm(1, point$mean[x], point$sd[x]),
mean = XPred$Est,
sd = sqrt(XPred$Sd ^ 2 + XPred$resError ^ 2),
log = FALSE
)))))
densitySD <- lapply(1:nrow(point), function(x)
apply((sapply(1:1000,
function(y)
dnorm(
rnorm(1, point$mean[x], point$sd[x]),
mean = XPred$Est,
sd = sqrt(XPred$Sd ^ 2 + XPred$resError ^ 2),
log = FALSE
))), 1, sd))
XPred$densitySd <- Reduce("*", densitySD)
} else {
density <- lapply(1:nrow(point), function(x)
dnorm(
point$mean[x],
mean = XPred$Est,
sd = sqrt(XPred$Sd ^ 2 + XPred$resError ^ 2),
log = TRUE)
)
XPred$densitySd <- 0
}
XPred$density <- exp(Reduce("+", density))
XPred$Est <- XPred$density
XPred$Sd <- XPred$densitySd
XPred$IntLower <- pmax(0, XPred$Est - 1.96 * XPred$Sd)
XPred$IntUpper <- XPred$Est + 1.96 * XPred$Sd
return(XPred)
}
combineSimilarityMaps <- function(XPredList,
normalize = FALSE,
normalType = "1",
weightProb = FALSE,
weightMap = NULL,
negZero = TRUE,
invWeight = FALSE,
weightDecay = 1) {
XPred <- XPredList[[1]]
XPred$density <- log(XPred$density)
if (length(XPredList) > 1) {
for (j in 2:length(XPredList)) {
closest <- sapply(1:nrow(XPred), function(x)
XPredList[[j]]$density[which.min((XPred$Longitude[x] - XPredList[[j]]$Longitude) ^ 2 +
(XPred$Latitude[x] - XPredList[[j]]$Latitude) ^ 2
)])
closestSD <- sapply(1:nrow(XPred), function(x)
XPredList[[j]]$densitySd[which.min((XPred$Longitude[x] - XPredList[[j]]$Longitude) ^ 2 +
(XPred$Latitude[x] - XPredList[[j]]$Latitude) ^ 2
)])
closestDist <- sapply(1:nrow(XPred), function(x)
min((XPred$Longitude[x] - XPredList[[j]]$Longitude) ^ 2 +
(XPred$Latitude[x] - XPredList[[j]]$Latitude) ^ 2
))
closest[closestDist > sqrt(100)] <- NA
closestSD[closestDist > sqrt(100)] <- NA
XPred$densitySd <- sqrt((XPred$densitySd ^ 2 + exp(XPred$density) ^ 2) * (closestSD ^ 2 + closest ^ 2) -
(exp(XPred$density) ^ 2) * (closest ^ 2))
XPred$density <- XPred$density + log(closest)
}
}
XPred$density <- exp(XPred$density)
if(!is.null(weightMap) & weightProb == TRUE){
closestWeights <- sapply(1:nrow(XPred), function(x)
weightMap$predictions$Est[which.min((XPred$Longitude[x] - weightMap$predictions$Longitude) ^ 2 +
(XPred$Latitude[x] - weightMap$predictions$Latitude) ^ 2
)])
closestWeightsOG <- closestWeights
if(!negZero){
closestWeights[closestWeightsOG<0] <- 0
}
if(invWeight){
closestWeights <- exp(-log(2)/weightDecay * closestWeights)
}
if(negZero){
closestWeights[closestWeightsOG<0] <- 0
}
XPred$density <- XPred$density * closestWeights
XPred$densitySd <- XPred$densitySd * sqrt(closestWeights)
}
XPred$Est <- XPred$density
if(normalize == TRUE){
if(normalType == "1"){
constant <- max(XPred$Est, na.rm = TRUE)
} else {
constant <- sum(XPred$Est, na.rm = TRUE) /
(diff(sort(unique(XPred$Longitude))[1:2]) *
diff(sort(unique(XPred$Latitude))[1:2]))
}
} else {
constant <- 1
}
XPred$Est <- XPred$Est / constant
XPred$Sd <- XPred$densitySd / sqrt(constant)
XPred$IntLower <- pmax(0, XPred$Est - 1.96 * XPred$Sd)
XPred$IntUpper <- XPred$Est + 1.96 * XPred$Sd
return(XPred)
}
createDifferenceMap <- function(XPred1, XPred2, operation = "-") {
if (inherits(XPred2, "numeric") && !inherits(XPred1, "numeric")) {
XPredNew <- XPred1
XPredNew$Est <- XPred2[1]
XPredNew$Sd <- XPred2[2]
XPred2 <- XPredNew
}
if (inherits(XPred1, "numeric") && !inherits(XPred2, "numeric")) {
XPredNew <- XPred2
XPredNew$Est <- XPred1[1]
XPredNew$Sd <- XPred1[2]
XPred1 <- XPredNew
}
if (inherits(XPred1, "numeric") && inherits(XPred2, "numeric")) {
lo <- seq(-180, 180, by = 0.5)
la <- seq(-90, 90, by = 0.5)
coord <- expand.grid(lo, la)
XPred1 <- data.frame(Est = XPred1[1], Sd = XPred1[2],
Longitude = coord[,1], Latitude = coord[,2])
XPred2 <- data.frame(Est = XPred2[1], Sd = XPred2[2],
Longitude = coord[,1], Latitude = coord[,2])
}
closestMean <- sapply(1:nrow(XPred1), function(x)
XPred2$Est[which.min((XPred1$Longitude[x] - XPred2$Longitude) ^ 2 +
(XPred1$Latitude[x] - XPred2$Latitude) ^ 2)])
closestSd <- sapply(1:nrow(XPred1), function(x)
XPred2$Sd[which.min((XPred1$Longitude[x] - XPred2$Longitude) ^ 2 +
(XPred1$Latitude[x] - XPred2$Latitude) ^ 2)])
closestDist <- sapply(1:nrow(XPred1), function(x)
min((XPred1$Longitude[x] - XPred2$Longitude) ^ 2 +
(XPred1$Latitude[x] - XPred2$Latitude) ^ 2
))
closestMean[closestDist > sqrt(2)] <- NA
closestSd[closestDist > sqrt(2)] <- NA
XPred <- XPred1
if(operation == "-"){
XPred$Est <- XPred1$Est - closestMean
XPred$Sd <- sqrt(XPred1$Sd ^ 2 + closestSd ^ 2)
}
if(operation == "+"){
XPred$Est <- XPred1$Est + closestMean
XPred$Sd <- sqrt(XPred1$Sd ^ 2 + closestSd ^ 2)
}
if(operation == "*"){
XPred$Est <- XPred1$Est * closestMean
XPred$Sd <- sqrt(XPred1$Sd ^ 2 * closestSd ^ 2 +
XPred1$Sd ^ 2 * closestMean ^ 2 +
closestSd ^ 2 * XPred1$Est ^ 2)
}
if(operation == "/"){
XPred$Est <- XPred1$Est / closestMean
XPred$Sd <- sqrt(((XPred1$Est ^ 2) / (closestMean ^ 2)) *
((XPred1$Sd ^ 2) / (XPred1$Est ^ 2) + (closestMean ^ 2) / (closestSd ^ 2)))
}
if(operation == "mean"){
XPred$Est <- (XPred1$Est + closestMean) / 2
XPred$Sd <- sqrt(XPred1$Sd ^ 2 + closestSd ^ 2) / sqrt(2)
}
if(operation == "weightedMean"){
VAR <- 1 / (1 / (XPred1$Sd ^ 2) + 1 / (closestSd ^ 2))
XPred$Est <- VAR * (XPred1$Est / (XPred1$Sd ^ 2) + closestMean / (closestSd ^ 2))
XPred$Sd <- sqrt(VAR)
}
if(operation == "weight"){
constant <- mean(closestMean)
closestMean <- closestMean / constant
closestSd <- closestSd / constant
XPred$Est <- XPred1$Est * closestMean
XPred$Sd <- sqrt(XPred1$Sd ^ 2 * closestSd ^ 2 +
XPred1$Sd ^ 2 * closestMean ^ 2 +
closestSd ^ 2 * XPred1$Est ^ 2)
}
if(operation == "pMin"){
XPred$Est <- pmin(XPred1$Est, closestMean, na.rm = T)
XPred1$Est[is.na(XPred1$Est)] <- -Inf
XPred$Sd <- sapply(1:length(XPred1$Sd), function(l) c(XPred1$Sd[l], closestSd[l])[which.min(c(XPred1$Est[l], closestMean[l]))])
XPred$Sd[is.na(XPred$Est)] <- NA
}
if(operation == "pMax"){
XPred$Est <- pmax(XPred1$Est, closestMean, na.rm = T)
XPred1$Est[is.na(XPred1$Est)] <- Inf
XPred$Sd <- sapply(1:length(XPred1$Sd), function(l) c(XPred1$Sd[l], closestSd[l])[which.max(c(XPred1$Est[l], closestMean[l]))])
XPred$Sd[is.na(XPred$Est)] <- NA
}
XPred$IntLower <- XPred$Est - 1.96 * XPred$Sd
XPred$IntUpper <- XPred$Est + 1.96 * XPred$Sd
return(XPred)
}
getMinima <- function(XPredPlot, nMin = 3, minDist = 250, minima = "Min"){
if(minima == "Max"){
XPredPlot$Est <- - 1* XPredPlot$Est
}
mins <- rep(NA, nMin)
mins[1] <- which.min(XPredPlot$Est)
y = 1
dist <- which(sqrt((XPredPlot$Longitude[mins[y]] - XPredPlot$Longitude) ^ 2 +
(XPredPlot$Latitude[mins[y]] - XPredPlot$Latitude) ^ 2) * 111 > minDist)
while(y < nMin & y < 9){
newMin <- which.min(XPredPlot$Est[dist])
newMinVal <- min(XPredPlot$Est[dist], na.rm = TRUE)
if(length(na.omit(newMin)) == 0){
return(na.omit(mins))
}
nearest <-
sqrt((XPredPlot$Longitude[dist][newMin] - XPredPlot$Longitude[dist]) ^ 2 +
(XPredPlot$Latitude[dist][newMin] - XPredPlot$Latitude[dist]) ^ 2)
neighbours <- XPredPlot$Est[dist][which(nearest <= sort(nearest)[5])]
if(newMinVal == min(neighbours, na.rm = TRUE)){
y = y +1
mins[y] = which(newMinVal == XPredPlot$Est)
dist <- intersect(dist, which(sqrt((XPredPlot$Longitude[mins[y]] - XPredPlot$Longitude) ^ 2 +
(XPredPlot$Latitude[mins[y]] - XPredPlot$Latitude) ^ 2) * 111 > minDist))
} else {
dist <- setdiff(dist, which(newMinVal == XPredPlot$Est))
}
if(length(na.omit(dist)) == 0){
return(na.omit(mins))
}
}
return(na.omit(mins))
}
#' Select Cluster Grouping
#'
#' @param data (data.frame) data if \code{cluster == TRUE} it must contain columns "spatial_cluster"
#' and "temporal_group"
#' @param cluster (logical) if TRUE, a column "cluster" is added to data
#' @param clusterResults (integer) either \code{0} or \code{1}, using "temporal_group" or
#' "spatial_cluster" for the column "cluster", respectively.
#'
#' @return (data.frame) data containing the column "cluster" if \code{cluster == TRUE}, else
#' just the input \code{data}
selectClusterGrouping <- function(data, cluster, clusterResults) {
if(cluster & !is.null(data$spatial_cluster)){
if(clusterResults == 0){
data$cluster <- data$temporal_group
} else {
data$cluster <- data$spatial_cluster
}
}
return(data)
}
#' Get Point Color
#'
#' @param data (data.frame) if cluster, different ids in the column "cluster" will receive a different colour
#' @param cluster (logical) if TRUE, return color for cluster
#' @param palName (character) name of the color palette
#' @param pColor single color or vector of colors that is returned if \code{cluster == FALSE}
#'
#' @return single color or vector of colors
getPColor <- function(data, cluster, palName, pColor) {
if (!cluster) return(pColor)
colorRampPalette(brewer.pal(8, palName))(max(data$cluster))[data$cluster]
}
Pandora-IsoMemo/iso-app documentation built on July 4, 2024, 7:07 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions getPColor selectClusterGrouping getMinima createDifferenceMap combineSimilarityMaps similarityMap createSimilarityMap roundUpNice plotTimeIntervals addFormattedAxis plotTimeCourse north.arrow getPlotRegion filled.contour2 plotDS plotMap3D plotMap
Documented in addFormattedAxis filled.contour2 getPColor plotDS plotMap plotMap3D plotTimeCourse selectClusterGrouping
#' Plots map of a spatial average or spread model from estimateMap() or estimateMapSpread() functions
#'
#' @param model return object of spatial or spread model
#' from estimateMap() or estimateMapSpread() functions
#' @param IndSelect for categorical model: selected category; shifts between categories in the center
#' @param arrow display north arrow TRUE/FALSE
#' @param scale display scale TRUE/FALSE
#' @param terrestrial show only estimates on land masses (1), oceans (-1) or all (0)
#' @param resolution spatial grid resolution of displayed (higher is slower but better quality)
#' @param interior show only convex hull TRUE/FALSE
#' @param grid show coordinate grid TRUE/FALSE
#' @param ncol number of colors for estimates
#' @param colors color scheme of estimates from RColorBrewer. defaults to "RdYlGn"
#' @param reverseColors inverse colour scheme
#' @param mapType Type of map. "Map" for simple estimates. "speed" for gradient
#' (mainly for spread model)
#' @param estType one of "Mean", "SE" and "Quantile". defaults to "Mean"
#' @param estQuantile quantile (only applicable if estType = "Quantile")
#' @param points show points on map TRUE/FALSE
#' @param pColor color of location marks drawn
#' @param colorsP color scale of location marks
#' @param pointShape pch shape of location marks
#' @param pointSize point size (if points = TRUE)
#' @param centerMap center of map, one of "Europe" and "Pacific"
#' @param StdErr maximum standard error to which estimates are displayed
#' @param rangex range of longitude values (x axis limits)
#' @param rangey range of latitude values (y axis limits)
#' @param mask boolean: Show output within range of points
#' @param maskRadius numeric: Show output within range of points in km
#' @param limitz restrict range of z (possible values "0-1", "0-100")
#' @param rangez range of estimated values (z axis limits)
#' @param dataCenter optional data.frame with two columns (longitude / latitude) for batch estimates
#' of a series of spatial points
#' @param RadiusBatch radius of batch spatial point estimates
#' @param textLabels text labels
#' @param pointLabels point size labels
#' @param pointColLabels point colour labels
#' @param fontSize font size
#' @param showScale show colour scale
#' @param setAxisLabels show axis/main/scale labels
#' @param mainLabel main label
#' @param yLabel y lab label
#' @param xLabel y lab label
#' @param scLabel scale lab label
#' @param northSize size of north arrow
#' @param scalSize size of scale
#' @param scaleX scale x orientation
#' @param scaleY scale y orientation
#' @param NorthX north arrow x orientation
#' @param NorthY north arrow y orientation
#' @param showModel show model
#' @param fontType font type
#' @param fontCol font color
#' @param titleMain show main title
#' @param titleScale show scale title
#' @param AxisSize axis title font size
#' @param AxisLSize axis label font size
#' @param cluster show clusters
#' @param clusterCol Cluster colors
#' @param pointDat data frame of points to add to plot
#' @param MinMax Min or Max, only for spread model
#' @param nMin Number of minima to compare, only for spread model
#' @param minDist Distance between minima/maxima, only for spread model
#' @param showMinOnMap Show minima on map yes/no, only for spread model if maptype = "Minima/Maxima"
#' @inheritParams filled.contour2
#'
#' @export
plotMap <- function(model,
IndSelect = NULL,
estType = "Mean", estQuantile = 0.9,
points = TRUE, pointSize = 1, StdErr = Inf,
rangex = range(model$data$Longitude, na.rm = TRUE),
rangey = range(model$data$Latitude, na.rm = TRUE),
rangez = NULL,
contourType = c("filled.contour", "contour"),
limitz = NULL, resolution = 100, interior = TRUE,
ncol = 10, colors = "RdYlGn", reverseColors = FALSE,
colorsP = NULL,
centerMap = "Europe",
pointShape = 4,
showScale = TRUE,
textLabels = NULL,
pointLabels = NULL,
pointColLabels = NULL,
fontSize = 11,
showModel = TRUE,
fontType = "sans",
fontCol = "black",
mask = FALSE,
maskRadius = 100,
pColor = "#000000",
dataCenter = NULL, RadiusBatch = NULL, terrestrial = 1,
grid = TRUE, arrow = TRUE, scale = TRUE, mapType = "Map",
titleMain = TRUE,
titleScale = TRUE,
setAxisLabels = FALSE,
mainLabel = "",
yLabel = "Latitude",
xLabel = "Longitude",
northSize = 0.2,
scalSize = 0.1,
scaleX = 0,
scaleY = 0.1,
NorthX = 0.025,
NorthY = 0.925,
scLabel = "",
AxisSize = 1,
AxisLSize = 1,
cluster = FALSE,
clusterCol = "Set1",
pointDat = NULL,
MinMax = "Min",
nMin = 3,
minDist = 250,
showMinOnMap = FALSE){
options(scipen=999)
contourType <- match.arg(contourType)
minRangeFactor <- 0.75
if((diff(rangex) / diff(rangey)) < minRangeFactor){
rangex[1] <- max(-180, mean(rangex) - minRangeFactor / 2 * diff(rangey))
rangex[2] <- min(180, mean(rangex) + minRangeFactor / 2 * diff(rangey))
}
minRangeFactor <- 0.5
if((diff(rangey) / diff(rangex)) < minRangeFactor){
rangey[1] <- max(-90, mean(rangey) - minRangeFactor / 2 * diff(rangex))
rangey[2] <- min(90, mean(rangey) + minRangeFactor / 2 * diff(rangex))
}
independent <- model$independent
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
if(IndSelect == "" | is.null(IndSelect)){
return(NULL)
}
model$model <- model$model[[IndSelect]]
}
if(is.null(rangez)){
rangez = range(model$data[, independent], na.rm = TRUE)
}
Bayes = TRUE
GAM = FALSE
if("gamm" %in% class(model$model)){
Bayes <- FALSE
}
if("kde" %in% class(model$model)){
GAM <- TRUE
independent <- "Density"
}
Maps <- loadMaps()
data <- model$data
if(!is.null(textLabels)){
textLabels <- textLabels[as.numeric(rownames(data)),1]
}
if(!is.null(pointLabels)){
pointLabels <- pointLabels[as.numeric(rownames(data)),1]
}
if(!is.null(pointColLabels)){
pointColLabels <- pointColLabels[as.numeric(rownames(data)),1]
}
sc <- model$sc
if (is.null(data)) return(NULL)
RadiusBatch <- RadiusBatch / 111
maskRadius <- maskRadius / 111
longitudes <- seq(max(-180, rangex[1] - 0.1 * diff(rangex)),
min(180, rangex[2] + 0.1 * diff(rangex)), length.out = resolution)
latitudes <- seq(max(-90, rangey[1] - 0.1 * diff(rangey)),
min(90, rangey[2] + 0.1 * diff(rangey)), length.out = resolution)
XPred <- expand.grid(Longitude = longitudes, Latitude = latitudes)
if(centerMap != "Europe"){
rangexEU <- rangex
rangex[rangexEU < 20] <- rangex[rangexEU < 20] + 160
rangex[rangexEU >= 20] <- (- 200 + rangex[rangexEU >= 20])
rangex <- rangex[order(rangex)]
longitudesPac <- longitudes
longitudes[longitudesPac < 20] <- longitudes[longitudesPac < 20] + 160
longitudes[longitudesPac >= 20] <- (- 200 + longitudes[longitudesPac >= 20])
XPredPac <- XPred
XPred$Longitude[XPredPac$Longitude < 20] <- XPred$Longitude[XPredPac$Longitude < 20] + 160
XPred$Longitude[XPredPac$Longitude >= 20] <- (- 200 + XPred$Longitude[XPredPac$Longitude >= 20])
dataPac <- data
dataPac$Longitude[data$Longitude < -20] <- dataPac$Longitude[data$Longitude < -20] + 200
dataPac$Longitude[data$Longitude >= -20] <- (- 160 + dataPac$Longitude[data$Longitude >= -20])
}
if (interior == TRUE){
if(centerMap != "Europe"){
poly.pnts <-
cbind(dataPac[chull(x = dataPac$Longitude, y = dataPac$Latitude), "Longitude"],
dataPac[chull(x = dataPac$Longitude, y = dataPac$Latitude), "Latitude"])
poly.pnts <- rbind(poly.pnts, poly.pnts[1, ])
draw <- point.in.polygon(XPredPac[, 1], XPredPac[, 2], poly.pnts[, 1], poly.pnts[, 2])
} else {
poly.pnts <-
cbind(data[chull(x = data$Longitude, y = data$Latitude), "Longitude"],
data[chull(x = data$Longitude, y = data$Latitude), "Latitude"])
poly.pnts <- rbind(poly.pnts, poly.pnts[1, ])
draw <- point.in.polygon(XPred[, 1], XPred[, 2], poly.pnts[, 1], poly.pnts[, 2])
}
}
if(mask == TRUE){
maskDraw <- extractMaskDraw(dat = centerPlotData(data, centerMap = centerMap),
maskRadius = maskRadius,
XPred = centerXPred(XPred = XPred,
XPredPac = XPredPac,
centerMap = centerMap))
}
if (!is.null(dataCenter)){
# this is batch mode
XPred <- do.call("rbind", lapply(1:nrow(dataCenter), function(x) {
longitudes <- seq(dataCenter[x, 1] - RadiusBatch,
dataCenter[x, 1] + RadiusBatch, length.out = 4)
latitudes <- seq(dataCenter[x, 2] - RadiusBatch,
dataCenter[x, 2] + RadiusBatch, length.out = 4)
XPred <- expand.grid(Longitude = longitudes, Latitude = latitudes)
cbind(XPred, id = x)
}))
}
XPred <- centerData(XPred, center = centerMap)
if (Bayes == TRUE & GAM == FALSE){
PredMatr <- Predict.matrix(sc, data = XPred)
betas <- model$model$beta
betaSigma <- model$model$betaSigma
Predictions <-
sapply(1:nrow(betas), function(x)
(PredMatr %*% betas[x, ]) * model$sRe + model$mRe)
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
Predictions <- invLogit(Predictions)
}
if(!is.null(betaSigma)){
PredMatrV <- Predict.matrix(model$scV, data = XPred)
PredictionsSigma <-
rowMeans(sqrt(sapply(1:nrow(betaSigma), function(x)
exp((PredMatrV %*% betaSigma[x, ])) / model$model$sigma[x]) * model$sRe^2))
} else {
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
PredictionsSigma <- sqrt(Predictions * (1-Predictions))
} else {
PredictionsSigma <- sqrt(mean(model$model$sigma) * model$sRe^2)
}
}
if(estType == "Mean"){
Est <- rowMeans(Predictions)
}
if(estType == "1 SE"){
Est <- apply(Predictions, 1, sd)
}
if(estType == "2 SE"){
Est <- apply(Predictions, 1, sd) * 2
}
if(estType == "1 SETOTAL"){
Est <- sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2)
}
if(estType == "2 SETOTAL"){
Est <- sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2) * 2
}
if(estType == "1 SD Population"){
Est <- PredictionsSigma * 1
}
if(estType == "2 SD Population"){
Est <- PredictionsSigma * 2
}
if(estType == "Quantile"){
Est <- apply(Predictions, 1, quantile, estQuantile)
}
if(estType == "QuantileTOTAL"){
Est <- rowMeans(Predictions + qnorm(estQuantile) * sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2))
}
XPred <- data.frame(XPred,
Est = Est,
Sd = apply(Predictions, 1, sd),
SDPop = PredictionsSigma,
SdTotal = sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2),
IntLower = apply(Predictions, 1, quantile, 0.025),
IntUpper = apply(Predictions, 1, quantile, 0.975),
IntLowerTotal = apply(Predictions + sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2),
1, quantile, 0.025),
IntUpperTotal = apply(Predictions + sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2),
1, quantile, 0.975),
resError = sqrt(mean(model$model$sigma + model$model$tau) * model$sRe^2))
}
if (Bayes == FALSE & GAM == FALSE){
Est <- predict(model$model$gam, XPred, se.fit = TRUE, type = "response")
if(estType == "1 SE"){
Est$fit <- Est$se.fit
}
if(estType == "2 SE"){
Est$fit <- Est$se.fit * 2
}
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
varM = Est$fit * (1-Est$fit)
} else {
varM = var(residuals(model$model$gam))
}
if(estType == "1 SETOTAL"){
Est$fit <- sqrt(Est$se.fit^2 + varM)
}
if(estType == "2 SETOTAL"){
Est$fit <- sqrt(Est$se.fit^2 + varM) * 2
}
if(estType == "1 SD Population"){
Est$fit <- sqrt(varM) * 1
}
if(estType == "2 SD Population"){
Est$fit <- sqrt(varM) * 2
}
if(estType == "Quantile"){
Est$fit <- Est$fit + qnorm(estQuantile) * Est$se.fit
}
if(estType == "QuantileTOTAL"){
Est$fit <- Est$fit + qnorm(estQuantile) *
sqrt(Est$se.fit^2 + varM)
}
XPred <- data.frame(XPred,
Est = Est$fit,
Sd = Est$se.fit,
SDPop = sqrt(varM),
SdTotal = sqrt(Est$se.fit^2 + varM),
IntLower = Est$fit - 1.96 * Est$se.fit,
IntUpper = Est$fit + 1.96 * Est$se.fit,
IntLowerTotal = Est$fit - 1.96 * sqrt(Est$se.fit^2 + varM),
IntUpperTotal = Est$fit + 1.96 * sqrt(Est$se.fit^2 + varM),
resError = sqrt(mean((predict(model$model$gam, type = "response") - model$model$gam$y)^2)))
}
if(GAM == TRUE){
Predictions <- sapply(1:length(model$model), function(x) predict(model$model[[x]], x = XPred[, 1:2]))
if(estType == "Mean"){
Est <- rowMeans(Predictions)
}
if(estType == "1 SE"){
Est <- apply(Predictions, 1, sd)
}
if(estType == "2 SE"){
Est <- apply(Predictions, 1, sd) * 2
}
if(estType == "Quantile"){
Est <- apply(Predictions, 1, quantile, estQuantile)
}
XPred <- data.frame(XPred,
Est = Est,
Sd = apply(Predictions, 1, sd),
IntLower = pmax(0, apply(Predictions, 1, quantile, 0.025)),
IntUpper = apply(Predictions, 1, quantile, 0.975))
}
if (estType != "1 SE" && estType != "2 SE" &&
estType != "1 SETOTAL" &&
estType != "2 SETOTAL" &&
!is.null(limitz) && !missing(limitz)){
if (limitz == "0-1"){
XPred$Est <- pmin(1, XPred$Est)
XPred$Est <- pmax(0, XPred$Est)
if(!is.null(XPred$IntLower)){
XPred$IntLower <- pmax(0, pmin(1, XPred$IntLower))
XPred$IntUpper <- pmax(0, pmin(1, XPred$IntUpper))
}
if(!is.null(XPred$IntLowerTotal)){
XPred$IntLowerTotal <- pmax(0, pmin(1, XPred$IntLowerTotal))
XPred$IntUpperTotal <- pmax(0, pmin(1, XPred$IntUpperTotal))
}
}
if (limitz == "0-100"){
XPred$Est <- pmin(100, XPred$Est)
XPred$Est <- pmax(0, XPred$Est)
if(!is.null(XPred$IntLower)){
XPred$IntLower <- pmax(0, pmin(100, XPred$IntLower))
XPred$IntUpper <- pmax(0, pmin(100, XPred$IntUpper))
}
if(!is.null(XPred$IntLowerTotal)){
XPred$IntLowerTotal <- pmax(0, pmin(100, XPred$IntLowerTotal))
XPred$IntUpperTotal <- pmax(0, pmin(100, XPred$IntUpperTotal))
}
}
}
if (!is.null(dataCenter)){
# this is batch mode
XPredCenter <- lapply(1:nrow(dataCenter), function(x){
XPred %>%
extractXPredCenter(centerX = dataCenter[x, 1],
centerY = dataCenter[x, 2],
Radius = RadiusBatch)
})
centerEstimates <- lapply(1:nrow(dataCenter), function(x){
XPredCenter[[x]] %>%
extractCenterEstimates(batch = TRUE)
})
meanCenter <- sapply(1:nrow(dataCenter), function(x) centerEstimates[[x]]$mean)
sdCenter <- sapply(1:nrow(dataCenter), function(x) centerEstimates[[x]]$sd)
IntLower <- sapply(1:nrow(dataCenter),
function(x) signif(mean(XPredCenter[[x]]$IntLower), 5))
IntUpper <- sapply(1:nrow(dataCenter),
function(x) signif(mean(XPredCenter[[x]]$IntUpper), 5))
if(is.null(XPredCenter[[1]]$SDPop)) {
dataCenter <- cbind(dataCenter, mean = meanCenter, sd = sdCenter,
IntLower = IntLower, IntUpper = IntUpper)
} else {
#population SD
SDPop <- sapply(1:nrow(dataCenter),
function(x) signif(sqrt(mean(XPredCenter[[x]]$SDPop, na.rm = TRUE)^2), 5))
sdCenterTotal <- sapply(1:nrow(dataCenter),
function(x) signif(sqrt(sum(sd(XPredCenter[[x]]$Est) ^ 2,
mean(XPredCenter[[x]]$Sd) ^ 2,
mean(XPredCenter[[x]]$SDPop) ^ 2,
na.rm = TRUE)), 5))
IntLowerTotal <- sapply(1:nrow(dataCenter),
function(x) signif(min(XPredCenter[[x]]$IntLowerTotal), 5))
IntUpperTotal <- sapply(1:nrow(dataCenter),
function(x) signif(max(XPredCenter[[x]]$IntUpperTotal), 5))
dataCenter <- cbind(dataCenter, mean = meanCenter, sd = sdCenter,
SDPop = SDPop,
sdTotal = sdCenterTotal,
IntLower = IntLower, IntUpper = IntUpper,
IntLowerTotal = IntLowerTotal, IntUpperTotal = IntUpperTotal)
}
return(dataCenter)
}
# keep $Est in XPred for later calculation of mean and sd
XPred$EstForCenter <- XPred$Est
if (interior == TRUE){
# this step can remove all $Est
XPred$Est[draw == 0] <- NA
}
if (mask == TRUE){
XPred$Est[maskDraw == 0] <- NA
}
if(GAM == TRUE){
XPred$Est[is.na(XPred$Est) | XPred$Est < 0] <- 0
}
XPred$Est[XPred$Sd > StdErr] <- NA
levels <- pretty(c(rangez[1], rangez[2]), n = ncol)
if(mapType == "Minima/Maxima"){
mins <- getMinima(XPred, nMin = nMin, minDist = minDist, minima = MinMax)
if(MinMax == "Min"){
probs <- round(as.vector(prop.table(table(factor(apply(Predictions[mins, ], 2, which.min), levels = 1:nMin)))), 3)
} else{
probs <- round(as.vector(prop.table(table(factor(apply(Predictions[mins, ], 2, which.max), levels = 1:nMin)))), 3)
}
dataMinPlot <- XPred[mins,]
dataMinPlotPred <- (t(Predictions[mins,]))
dataMinPlotPred <- data.frame(name = rep(paste0(MinMax, "ima ", 1:nMin,
"\n Coord. = (",round(dataMinPlot$Latitude, 2),", ",
round(dataMinPlot$Longitude, 2),") \n Probability = ", prob = probs),
each = nrow(dataMinPlotPred)),
estimate = as.vector(dataMinPlotPred))
minMeans <- aggregate(dataMinPlotPred$estimate, by=list(name=dataMinPlotPred$name), FUN=mean)
minMedians <- aggregate(dataMinPlotPred$estimate, by=list(name=dataMinPlotPred$name), FUN=median)
minq68 <- aggregate(dataMinPlotPred$estimate, by=list(name=dataMinPlotPred$name), FUN=quantile, 0.68)
minq32 <- aggregate(dataMinPlotPred$estimate, by=list(name=dataMinPlotPred$name), FUN=quantile, 0.32)
minq95 <- aggregate(dataMinPlotPred$estimate, by=list(name=dataMinPlotPred$name), FUN=quantile, 0.975)
minq05 <- aggregate(dataMinPlotPred$estimate, by=list(name=dataMinPlotPred$name), FUN=quantile, 0.025)
meanEst <- q32 <- q68 <- q95 <- q05 <- NULL
dataSummary <- data.frame(name = unique(dataMinPlotPred$name), meanEst = minMeans[,2], median = minMedians[,2],
q68 = minq68[,2], q32 = minq32[,2],
q95 = minq95[,2], q05 = minq05[,2])
if(showMinOnMap == "1"){
g <- ggplot(dataSummary, aes_(x = ~name, fill = ~name)) + theme_light()
g <- g + geom_boxplot(
mapping = aes(
lower = q32,
upper = q68,
middle = median,
ymin = q05,
ymax = q95
),
stat = "identity"
) + geom_errorbar(aes(ymin = meanEst, ymax = meanEst), linetype = "dashed", data = dataSummary)+ theme(legend.position="none") +
labs(title=paste0("Comparison of local ", MinMax, "ima")) + xlab("")
print(g)
return(list(XPred = XPred))
}
}
#if (!all(is.na(XPred$Est))){
if (mapType == "Speed"){
if(centerMap != "Europe"){
longitudes <- longitudesPac[order(longitudesPac)]
latitudes <- latitudes[order(latitudes, longitudesPac)]
XPredPlot <- data.frame(XPredPac[order(XPredPac$Latitude, XPredPac$Longitude),],
Est = XPred$Est[order(XPredPac$Latitude, XPredPac$Longitude)])
rangex <- rangexEU
} else {
XPredPlot <- XPred
}
z = matrix(XPredPlot$Est, ncol = resolution)
test.rast <- raster(ncol= nrow(z), nrow = nrow(z), xmn = 0, xmx = 1, ymn = 0, ymx = 1)
test.rast[] <- as.vector(z)
z2 <- abs(matrix(getValues(terrain(test.rast, unit = "tangent")), ncol = nrow(z)))
tankilometers <- sqrt(diff(latitudes)[1] * diff(longitudes)[1]) * 111
z2 <- log(((1/z2) / tankilometers) + 1E-10)
z2levels <- signif(exp(pretty(as.vector(z2), n = ncol)), 2)
q99 <- quantile(as.vector(exp(z2)), na.rm = TRUE, 0.975)
z2levels <- z2levels[z2levels < q99]
z2levels <- unique(c(z2levels, signif(q99, 2),
signif(max(as.vector(exp(z2)), na.rm = TRUE), 2)))
# z2levels <- signif(pretty(c((rangez[1])^0.05, (rangez[2])^0.05), n = ncol)^20, 2)
# z2levels[1] <- rangez[1]
# z2levels[length(z2levels)] <- rangez[2]
# if(length(unique(z2levels)) < 4){
# z2levels <- signif(pretty(c((1)^0.05, (1000)^0.05), n = ncol)^20, 2)
# }
colors <- colorRampPalette(brewer.pal(9, colors))(length(z2levels) - 1)
if(reverseColors){
colors <- rev(colors)
}
levelsLegend <- z2levels
if(!showModel){
z2 <- rep(NA, length(z2))
}
if(titleMain){
main = ""
} else {
if(setAxisLabels){
main = mainLabel
} else {
main = independent
}
}
if(titleScale){
mainS = ""
} else {
if(setAxisLabels){
mainS = scLabel
} else {
mainS = "km/time unit"
}
}
if(setAxisLabels){
xlab = xLabel
ylab = yLabel
} else {
xlab = "Longitude"
ylab = "Latitude"
}
#for geotiff export
if(centerMap != "Europe"){
XPred$Est <- exp(as.vector(z2))[order(XPred$Latitude, XPred$Longitude)]
} else {
XPred$Est <- exp(as.vector(z2))
}
filled.contour2(longitudes, latitudes, z = z2,
contourType = contourType,
cex.axis = 1.5, cex.main = 1.5, cex.lab = 1.5,
xlim = rangex, ylim = rangey,
zlim = range(z2, finite = TRUE),
levels = log(z2levels + 1E-10), nlevels = length(z2levels),
col = colors,
asp = 1,
showScale = showScale,
key.axes = axis(4, at = log(levelsLegend + 1E-10), labels = levelsLegend),
key.title = title(main = mainS, cex.main = 0.8),
plot.title = {title(cex.lab = AxisSize, xlab = xlab, ylab = ylab, main = main)},
plot.axes = {
par(fg = "black", col="black");
if (terrestrial == 1){
if(centerMap != "Europe"){
sp::plot(Maps$ocean160, add = T, col = "lightblue", lwd = 1, border = NA);
sp::plot(Maps$ocean200, add = T, col = "lightblue", lwd = 1, border = NA);
} else {
sp::plot(Maps$ocean, add = T, col = "lightblue", lwd = 1);
}
}
if (terrestrial == -1){
if(centerMap != "Europe"){
sp::plot(Maps$land160, add = T, lwd = 1, col = "grey96", border = NA);
sp::plot(Maps$land200, add = T, lwd = 1, col = "grey96", border = NA);
} else {
sp::plot(Maps$land, add = T, lwd = 1, col = "grey96", border = NA);
}
}
if(centerMap != "Europe"){
sp::plot(Maps$coast160, add = T, lwd = 1);
sp::plot(Maps$coast200, add = T, lwd = 1);
} else {
sp::plot(Maps$coast, add = T, lwd = 1);
}
if (grid == TRUE){
if(centerMap != "Europe"){
sp::plot(Maps$grids160, add = T, col = "grey", lty = 2, xlim = c(0, 1));
sp::plot(Maps$grids200, add = T, col = "grey", lty = 2, xlim = c(0, 1));
} else {
sp::plot(Maps$grids, add = T, col = "grey", lty = 2, xlim = c(0, 1));
}
}
if(centerMap != "Europe"){
sp::plot(Maps$borders160, add = T, col = "darkgrey", lwd = 1);
sp::plot(Maps$borders200, add = T, col = "darkgrey", lwd = 1);
} else {
sp::plot(Maps$borders, add = T, col = "darkgrey", lwd = 1);
}
if (points == TRUE){
if(!is.null(pointLabels)){
pointLabels <- as.numeric(pointLabels)
pointLabels <- pointLabels - min(pointLabels, na.rm = TRUE)
pointLabels <- pointLabels / max(pointLabels, na.rm = TRUE)
pointSize <- (pointSize * pointLabels + 0.1) * 2
}
if(!is.null(pointColLabels)){
pointColLabels <- as.numeric(pointColLabels)
pointColLabels <- pointColLabels - min(pointColLabels, na.rm = TRUE)
pointColLabels <- pointColLabels / max(pointColLabels, na.rm = TRUE)
pColor <- (colorRampPalette(brewer.pal(9, colorsP))(101))[round(pointColLabels, 2) * 100]
}
if(cluster & !is.null(data$spatial_cluster)){
pColor <- colorRampPalette(brewer.pal(8, clusterCol))(max(data$spatial_cluster, na.rm=TRUE))[data$spatial_cluster]
if((!"long_temporal_group_reference_point" %in% names(data))){
data_names <- c("spatial_cluster", "long_centroid_spatial_cluster", "lat_centroid_spatial_cluster")
} else {
data_names <- c("temporal_group", "long_temporal_group_reference_point", "lat_temporal_group_reference_point")
}
centroids <- unique(data[, data_names])
centroids <- centroids[order(centroids[,1]), ]
if(centerMap != "Europe"){
centroids2 <- centroids
centroids2[,2][centroids[,2] < -20] <- centroids2[,2][centroids[,2] < -20] + 200
centroids2[,2][centroids[,2] >= -20] <- (- 160 + centroids2[,2][centroids[,2] >= -20])
centroids <- centroids2
}
}
if(centerMap != "Europe"){
points(dataPac$Latitude ~ dataPac$Longitude,
col = pColor, lwd = 2,
pch = pointShape, cex = pointSize);
} else {
points(data$Latitude ~ data$Longitude,
col = pColor, lwd = 2,
pch = pointShape, cex = pointSize);
}
if(cluster & !is.null(data$spatial_cluster)){
points(centroids[, 2:3], lwd = 2,
pch = pointShape, cex = pointSize * 2.5, col = colorRampPalette(brewer.pal(8, clusterCol))(max(data$spatial_cluster, na.rm=TRUE)))
text(centroids[, 2:3], labels = paste0("Cluster_", centroids[,1]), pos = 4,
cex = fontSize * 1.5, col = fontCol, family = fontType)
}
}
if(!is.null(textLabels)){
if(centerMap != "Europe"){
text(dataPac$Latitude ~ dataPac$Longitude,
labels = as.character(textLabels), pos = 4,
cex = fontSize, col = fontCol, family = fontType)
} else {
text(data$Latitude ~ data$Longitude,
labels = as.character(textLabels), pos = 4,
cex = fontSize, col = fontCol, family = fontType)
}
}
if(centerMap != "Europe"){
lab <- pretty(rangex)
lab[pretty(rangex) >= 20] <- lab[pretty(rangex) >= 20] - 200
lab[pretty(rangex) < 20] <- lab[pretty(rangex) < 20] + 160
axis(1, at = pretty(rangex), labels = lab, cex.axis = AxisLSize);
} else{
axis(1, cex.axis = AxisLSize);
}
axis(2, cex.axis = AxisLSize);
}
)
if (arrow == TRUE){
if(showScale == TRUE){
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
} else {
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
}
}
if (scale == TRUE){
if(showScale == TRUE){
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
} else {
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
}
}
return(list(XPred = XPred))
}
if(GAM == TRUE){
colors <- c("#FFFFFF", colorRampPalette(brewer.pal(9, colors))(length(levels) - 2))
} else{
colors <- colorRampPalette(brewer.pal(9, colors))(length(levels) - 1)
}
if(reverseColors){
colors <- rev(colors)
}
levelsLegend <- levels
if(length(levels) > 25){
par(fg = NA, col="black")
levelsLegend <- pretty(c(rangez[1], rangez[2]),
n = pmin(20, ceiling(ncol / 2)))
}
cex4 <- 1
if(max(abs(XPred$Est), na.rm = TRUE) > 9999 | max(abs(XPred$Est), na.rm = TRUE) < 0.05){
cex4 <- 0.7
}
if(centerMap != "Europe"){
longitudes <- longitudesPac[order(longitudesPac)]
latitudes <- latitudes[order(latitudes, longitudesPac)]
XPredPlot <- data.frame(XPredPac[order(XPredPac$Latitude, XPredPac$Longitude),],
Est = XPred$Est[order(XPredPac$Latitude, XPredPac$Longitude)])
rangex <- rangexEU
} else {
XPredPlot <- XPred
}
if(!showModel){
XPredPlot$Est <- rep(NA, length(XPredPlot$Est))
}
if(titleMain){
main = ""
} else {
if(setAxisLabels){
main = mainLabel
} else {
main = independent
}
}
if(titleScale){
mainS = ""
} else {
if(setAxisLabels){
mainS = scLabel
} else {
mainS = independent
}
}
if(setAxisLabels){
xlab = xLabel
ylab = yLabel
} else {
xlab = "Longitude"
ylab = "Latitude"
}
filled.contour2(longitudes, latitudes, z = matrix(XPredPlot$Est, ncol = resolution),
contourType = contourType,
xlim = rangex, ylim = rangey, levels = levels, zlim = rangez,
col = colors,
showScale = showScale,
cex.axis = 1.5, cex.main = 1.5, cex.lab = 1.5,
asp = 1, key.axes = axis(side = 4, at = levelsLegend, tick = FALSE, cex.axis = cex4),
key.title = title(main = mainS, cex.main = 0.8),
plot.title = {title(cex.lab = AxisSize, xlab = xlab, ylab = ylab, main = main)},
plot.axes = {
par(fg = "black", col="black");
if (terrestrial == 1){
if(centerMap != "Europe"){
sp::plot(Maps$ocean160, add = T, col = "lightblue", lwd = 1, border = NA);
sp::plot(Maps$ocean200, add = T, col = "lightblue", lwd = 1, border = NA);
} else {
sp::plot(Maps$ocean, add = T, col = "lightblue", lwd = 1);
}
}
if (terrestrial == -1){
if(centerMap != "Europe"){
sp::plot(Maps$land160, add = T, lwd = 1, col = "grey96", border = NA);
sp::plot(Maps$land200, add = T, lwd = 1, col = "grey96", border = NA);
} else {
sp::plot(Maps$land, add = T, lwd = 1, col = "grey96", border = NA);
}
}
if(centerMap != "Europe"){
sp::plot(Maps$coast160, add = T, lwd = 1);
sp::plot(Maps$coast200, add = T, lwd = 1);
} else {
sp::plot(Maps$coast, add = T, lwd = 1);
}
if (grid == TRUE){
if(centerMap != "Europe"){
sp::plot(Maps$grids160, add = T, col = "grey", lty = 2, xlim = c(0, 1));
sp::plot(Maps$grids200, add = T, col = "grey", lty = 2, xlim = c(0, 1));
} else {
sp::plot(Maps$grids, add = T, col = "grey", lty = 2, xlim = c(0, 1));
}
}
if(centerMap != "Europe"){
sp::plot(Maps$borders160, add = T, col = "darkgrey", lwd = 1);
sp::plot(Maps$borders200, add = T, col = "darkgrey", lwd = 1);
} else {
sp::plot(Maps$borders, add = T, col = "darkgrey", lwd = 1);
}
if (points == TRUE){
if(!is.null(pointLabels)){
pointLabels <- as.numeric(pointLabels)
pointLabels <- pointLabels - min(pointLabels, na.rm = TRUE)
pointLabels <- pointLabels / max(pointLabels, na.rm = TRUE)
pointSize <- (pointSize * pointLabels + 0.1) * 2
}
if(!is.null(pointColLabels)){
pointColLabels <- as.numeric(pointColLabels)
pointColLabels <- pointColLabels - min(pointColLabels, na.rm = TRUE)
pointColLabels <- pointColLabels / max(pointColLabels, na.rm = TRUE)
pColor <- (colorRampPalette(brewer.pal(9, colorsP))(101))[round(pointColLabels, 2) * 100]
}
if(cluster & !is.null(data$spatial_cluster)){
pColor <- colorRampPalette(brewer.pal(8, clusterCol))(max(data$spatial_cluster, na.rm=TRUE))[data$spatial_cluster]
if((!"long_temporal_group_reference_point" %in% names(data))){
data_names <- c("spatial_cluster", "long_centroid_spatial_cluster", "lat_centroid_spatial_cluster")
} else {
data_names <- c("temporal_group", "long_temporal_group_reference_point", "lat_temporal_group_reference_point")
}
centroids <- unique(data[, data_names])
centroids <- centroids[order(centroids[,1]), ]
if(centerMap != "Europe"){
centroids2 <- centroids
centroids2[,2][centroids[,2] < -20] <- centroids2[,2][centroids[,2] < -20] + 200
centroids2[,2][centroids[,2] >= -20] <- (- 160 + centroids2[,2][centroids[,2] >= -20])
centroids <- centroids2
}
}
if(centerMap != "Europe"){
points(dataPac$Latitude ~ dataPac$Longitude,
col = pColor, lwd = 2,
pch = pointShape, cex = pointSize);
} else {
points(data$Latitude ~ data$Longitude,
col = pColor, lwd = 2,
pch = pointShape, cex = pointSize);
}
if(cluster & !is.null(data$spatial_cluster)){
points(centroids[, 2:3], lwd = 2,
pch = pointShape, cex = pointSize * 2.5, col = colorRampPalette(brewer.pal(8, clusterCol))(max(data$spatial_cluster, na.rm=TRUE)))
text(centroids[, 2:3], labels = paste0("Cluster_", centroids[,1]), pos = 4,
cex = fontSize * 1.5, col = fontCol, family = fontType)
}
}
if(mapType == "Minima/Maxima"){
points(dataMinPlot$Latitude ~ dataMinPlot$Longitude, lwd = 2, col = pColor,
pch = pointShape, cex = pointSize * 2.5)
text(dataMinPlot$Latitude ~ dataMinPlot$Longitude, labels = paste0(MinMax, "ima", 1:nMin), pos = 4,
cex = fontSize * 1.5, col = fontCol, family = fontType)
}
if(!is.null(textLabels)){
if(centerMap != "Europe"){
text(dataPac$Latitude ~ dataPac$Longitude,
labels = as.character(textLabels), pos = 4,
cex = fontSize, col = fontCol, family = fontType)
} else {
text(data$Latitude ~ data$Longitude,
labels = as.character(textLabels), pos = 4,
cex = fontSize, col = fontCol, family = fontType)
}
}
if(!is.null(pointDat) & NROW(pointDat) > 0){
points(x = pointDat$x, y = pointDat$y, cex = pointDat$pointSize, col = pointDat$pointColor, pch = 16)
text(pointDat$y ~ pointDat$x, labels = pointDat$label, pos = 4, cex = 1.75)
}
if(centerMap != "Europe"){
lab <- pretty(rangex)
lab[pretty(rangex) >= 20] <- lab[pretty(rangex) >= 20] - 200
lab[pretty(rangex) < 20] <- lab[pretty(rangex) < 20] + 160
axis(1, at = pretty(rangex), labels = lab, cex.axis = AxisLSize);
} else{
axis(1, cex.axis = AxisLSize);
}
axis(2, cex.axis = AxisLSize);
}
)
# } else {
# plot(1:10, xlim = range(data$Longitude), ylim = range(data$Latitude),
# xlab = "Longitude", ylab = "Latitude")
# plot(Maps$ocean, add = T, col = "lightblue", lwd = 1)
# sp::plot(Maps$grids, add = T, col = "grey", lty = 2, xlim = c(0, 1))
# sp::plot(Maps$borders, add = T, col = "darkgrey", lwd = 1)
# text(mean(data$Longitude),mean(data$Latitude), cex = 1,
# "No estimates to plot, please adjust \"Display up to max standard error\" slider",
# col = "red")
# }
if (arrow == TRUE){
if(showScale == TRUE){
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
} else {
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
}
}
if (scale == TRUE){
if(showScale == TRUE){
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
} else {
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
}
}
return(list(XPred = XPred))
}
#' Plots time slice map of a spatio-temporal model from estimateMap3D() function
#'
#' @param model return object of a spatio-temporal model from estimateMap3D() function
#' @param IndSelect for categorical model: selected category
#' @param time time slice value for map
#' @param arrow display north arrow TRUE/FALSE
#' @param scale display scale TRUE/FALSE
#' @param terrestrial show only estimates on land masses (1), oceans (-1) or all (0)
#' @param resolution spatial grid resolution of displayed (higher is slower but better quality)
#' @param interior show only convex hull 0 "none", 1 "3D", 2 "time sliced 2D"
#' @param addU numeric: years of added uncertainty for time sliced 2D-convex hull
#' @param grid show coordinate grid TRUE/FALSE
#' @param ncol number of colors for estimates
#' @param pColor color of location marks drawn
#' @param colorsP color scale of location marks
#' @param pointShape pch shape of location marks
#' @param colors color scheme of estimates from RColorBrewer. defaults to "RdYlGn"
#' @param reverseColors inverse colour scheme
#' @param estType one of "Mean", "SE" and "Quantile". defaults to "Mean"
#' @param estQuantile quantile (only applicable if estType = "Quantile")
#' @param points show points on map TRUE/FALSE
#' @param pointSize point size (if points = TRUE)
#' @param StdErr maximum standard error to which estimates are displayed
#' @param rangex range of longitude values (x axis limits)
#' @param rangey range of latitude values (y axis limits)
#' @param mask boolean: Show output within range of points
#' @param maskRadius numeric: Show output within range of points in km
#' @param limitz restrict range of z (possible values "0-1", "0-100")
#' @param rangez range of estimated values (z axis limits)
#' @param centerMap center of map, one of "Europe" and "Pacific"
#' @param dataCenter optional data.frame with three columns (longitude / latitude / time)
#' @param RadiusBatch radius of batch spatial point estimates
#' @param textLabels text labels
#' @param pointLabels point size labels
#' @param pointColLabels point colour labels
#' @param fontSize font size
#' @param showModel show model
#' @param fontType font type
#' @param fontCol font color
#' @param titleMain show main title
#' @param titleScale show scale title
#' @param setAxisLabels show axis/main/scale labels
#' @param mainLabel main label
#' @param yLabel y lab label
#' @param xLabel y lab label
#' @param northSize size of north arrow
#' @param scalSize size of scale
#' @param scaleX scale x orientation
#' @param scaleY scale y orientation
#' @param NorthX north arrow x orientation
#' @param NorthY north arrow y orientation
#' @param scLabel scale lab label
#' @param AxisSize axis title font size
#' @param AxisLSize axis label font size
#' @param cluster show clusters
#' @param clusterAll show all cluster points
#' @param clusterResults temporal groups or spatial clusters
#' @param clusterCol Cluster colors
#' @param pointDat data frame of points to add to plot
#' @param plotRetNull return predictions
#' @inheritParams filled.contour2
#'
#' @export
plotMap3D <- function(model,
time,
IndSelect = NULL,
estType = "Mean", estQuantile = 0.9,
points = TRUE, pointSize = 1, StdErr = Inf,
rangex = range(model$data$Longitude, na.rm = TRUE),
rangey = range(model$data$Latitude, na.rm = TRUE),
rangez = NULL,
contourType = c("filled.contour", "contour"),
limitz = "none",
centerMap = "Europe",
addU = 0,
pointShape = 4,
pColor = "#000000",
colorsP = NULL,
showModel = TRUE,
textLabels = NULL,
pointLabels = NULL,
pointColLabels = NULL,
fontSize = 11,
mask = FALSE,
maskRadius = 100,
showScale = TRUE,
fontType = "sans",
fontCol = "black",
resolution = 100, interior = TRUE,
ncol = 10, colors = "RdYlGn", reverseColors = FALSE,
dataCenter = NULL,
RadiusBatch = 100, terrestrial = 1,
grid = TRUE, arrow = TRUE, scale = TRUE,
titleMain = TRUE,
titleScale = TRUE,
setAxisLabels = FALSE,
mainLabel = "",
yLabel = "Latitude",
xLabel = "Longitude",
scLabel = "",
northSize = 0.2,
scalSize = 0.1,
scaleX = 0,
scaleY = 0.1,
NorthX = 0.025,
NorthY = 0.925,
AxisSize = 1,
AxisLSize = 1,
cluster = FALSE,
clusterAll = "0",
clusterResults = 0,
clusterCol = "Set1",
pointDat = NULL,
plotRetNull = FALSE){
options(scipen = 999)
contourType <- match.arg(contourType)
minRangeFactor <- 0.75
if((diff(rangex) / diff(rangey)) < minRangeFactor){
rangex[1] <- max(-180, mean(rangex) - minRangeFactor / 2 * diff(rangey))
rangex[2] <- min(180, mean(rangex) + minRangeFactor / 2 * diff(rangey))
}
minRangeFactor <- 0.5
if((diff(rangey) / diff(rangex)) < minRangeFactor){
rangey[1] <- max(-90, mean(rangey) - minRangeFactor / 2 * diff(rangex))
rangey[2] <- min(90, mean(rangey) + minRangeFactor / 2 * diff(rangex))
}
independent <- model$independent
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
if(is.null(IndSelect) || IndSelect == ""){
return(NULL)
}
model$model <- model$model[[IndSelect]]
}
if(is.null(rangez)){
rangez = range(model$data[, independent], na.rm = TRUE)
}
Bayes = TRUE
GAM = FALSE
if("gamm" %in% class(model$model)){
Bayes = FALSE
}
if("kde" %in% class(model$model)){
GAM = TRUE
independent = "Density"
}
Maps <- loadMaps()
data <- model$data
sc <- model$sc
if(!is.null(textLabels)){
textLabels <- textLabels[as.numeric(rownames(data)),1]
}
if(!is.null(pointLabels)){
pointLabels <- pointLabels[as.numeric(rownames(data)),1]
}
if(!is.null(pointColLabels)){
pointColLabels <- pointColLabels[as.numeric(rownames(data)),1]
}
if (is.null(data)) return(NULL)
RadiusBatch <- RadiusBatch / 111
maskRadius <- maskRadius / 300
longitudes <- seq(max(-180, rangex[1] - 0.1 * diff(rangex)),
min(180, rangex[2] + 0.1 * diff(rangex)), length.out = resolution)
latitudes <- seq(max(-90, rangey[1] - 0.1 * diff(rangey)),
min(90, rangey[2] + 0.1 * diff(rangey)), length.out = resolution)
XPred <- cbind(expand.grid(Longitude2 = (longitudes - mean(data$Longitude)) / sd(data$Longitude),
Latitude2 = (latitudes - mean(data$Latitude)) / sd(data$Latitude)),
expand.grid(Longitude = longitudes, Latitude = latitudes), Date = time)
if(centerMap != "Europe"){
rangexEU <- rangex
rangex[rangexEU < 20] <- rangex[rangexEU < 20] + 160
rangex[rangexEU >= 20] <- (- 200 + rangex[rangexEU >= 20])
rangex <- rangex[order(rangex)]
longitudesPac <- longitudes
longitudes[longitudesPac < 20] <- longitudes[longitudesPac < 20] + 160
longitudes[longitudesPac >= 20] <- (- 200 + longitudes[longitudesPac >= 20])
XPredPac <- XPred
XPred$Longitude[XPredPac$Longitude < 20] <- XPred$Longitude[XPredPac$Longitude < 20] + 160
XPred$Longitude[XPredPac$Longitude >= 20] <- (- 200 + XPred$Longitude[XPredPac$Longitude >= 20])
XPred$Longitude2 = (XPred$Longitude - mean(data$Longitude)) / sd(data$Longitude)
dataPac <- data %>%
centerPlotData(centerMap = centerMap)
dataTPac <- dataPac %>%
filterT(addU = addU, time = time)
}
if (interior > 0){
if(centerMap != "Europe"){
if(interior == 1){
cData <- rbind(data.frame(Date = dataPac$Date + 2 * dataPac$Uncertainty, dataPac[, c("Longitude", "Latitude")]),
data.frame(Date = dataPac$Date - 2 * dataPac$Uncertainty, dataPac[, c("Longitude", "Latitude")]))
if(nrow(cData) > 0){
convHull <- convhulln(cData)
draw <- inhulln(convHull, as.matrix(XPredPac[, c(5,3,4)]))
} else {
draw <- rep(0, nrow(XPred))
}
} else {
cData <- dataTPac[, c("Longitude", "Latitude")]
if(nrow(cData) > 2){
convHull <- convhulln(cData)
draw <- inhulln(convHull, as.matrix(XPredPac[, c(3,4)]))
} else {
draw <- rep(0, nrow(XPred))
}
}
} else {
if(interior == 1){
cData <- rbind(data.frame(Date = data$Date + 2 * data$Uncertainty, data[, c("Longitude", "Latitude")]),
data.frame(Date = data$Date - 2 * data$Uncertainty, data[, c("Longitude", "Latitude")]))
if(nrow(unique(cData)) > 3){
convHull <- convhulln(cData)
draw <- inhulln(convHull, as.matrix(XPred[, c(5,3,4)]))
} else {
draw <- rep(0, nrow(XPred))
}
} else {
cData <- filterT(data, addU = addU, time = time)[, c("Longitude", "Latitude")]
if(nrow(unique(cData)) > 3){
convHull <- convhulln(cData)
draw <- inhulln(convHull, as.matrix(XPred[, c(3,4)]))
} else {
draw <- rep(0, nrow(XPred))
}
}
}
}
if(mask == TRUE){
if(exists("cData")){
maskData <- unique(cData[, c("Longitude", "Latitude")])
} else {
maskData <- data
}
maskDraw <- extractMaskDraw(dat = maskData,
maskRadius = maskRadius,
XPred = centerXPred(XPred = XPred,
XPredPac = XPredPac,
centerMap = centerMap))
}
if (!is.null(dataCenter)){
# this is batch mode
XPred <- do.call("rbind", lapply(1:nrow(dataCenter), function(x) {
longitudes <- seq(dataCenter[x, 1] - RadiusBatch,
dataCenter[x, 1] + RadiusBatch, length.out = 4)
latitudes <- seq(dataCenter[x, 2] - RadiusBatch,
dataCenter[x, 2] + RadiusBatch, length.out = 4)
XPred <- expand.grid(Longitude2 = (longitudes - mean(data$Longitude)) / sd(data$Longitude),
Latitude2 = (latitudes - mean(data$Latitude)) / sd(data$Latitude),
time = dataCenter[x, 3])
XPred$Longitude <- XPred$Longitude2 * sd(data$Longitude) + mean(data$Longitude)
XPred$Latitude <- XPred$Latitude2 * sd(data$Latitude) + mean(data$Latitude)
cbind(XPred, id = x)
}))
}
if (Bayes == TRUE & GAM == FALSE){
PredMatr <- Predict.matrix(sc,
data = data.frame(XPred,
Date2 = (time - mean(data$Date)) /
sd(data$Date)))
betas <- model$model$beta
betaSigma <- model$model$betaSigma
Predictions <-
sapply(1:nrow(betas), function(x)
PredMatr %*% betas[x, ] * model$sRe + model$mRe)
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
Predictions <- invLogit(Predictions)
}
if(!is.null(betaSigma)){
PredMatrV <- Predict.matrix(model$scV,
data = data.frame(XPred,
Date2 = (time - mean(data$Date)) /
sd(data$Date)))
PredictionsSigma <-
rowMeans(sqrt(sapply(1:nrow(betaSigma), function(x)
exp((PredMatrV %*% betaSigma[x, ])) / model$model$sigma[x]) * model$sRe^2))
} else {
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
PredictionsSigma <- sqrt(Predictions * (1-Predictions))
} else {
PredictionsSigma <- sqrt(mean(model$model$sigma) * model$sRe^2)
}
}
if(estType == "Mean"){
Est <- rowMeans(Predictions)
}
if(estType == "1 SE"){
Est <- apply(Predictions, 1, sd)
}
if(estType == "2 SE"){
Est <- apply(Predictions, 1, sd) * 2
}
if(estType == "1 SETOTAL"){
Est <- sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2)
}
if(estType == "2 SETOTAL"){
Est <- sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2) * 2
}
if(estType == "1 SD Population"){
Est <- PredictionsSigma * 1
}
if(estType == "2 SD Population"){
Est <- PredictionsSigma * 2
}
if(estType == "Quantile"){
Est <- apply(Predictions, 1, quantile, estQuantile)
}
if(estType == "QuantileTOTAL"){
Est <- rowMeans(Predictions + qnorm(estQuantile) * sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2))
}
XPred <-
data.frame(XPred,
Est = Est,
Sd = apply(Predictions, 1, sd),
SDPop = PredictionsSigma,
SdTotal = sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2),
IntLower = apply(Predictions, 1, quantile, 0.025),
IntUpper = apply(Predictions, 1, quantile, 0.975),
IntLowerTotal = apply(Predictions + sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2), 1, quantile, 0.025),
IntUpperTotal = apply(Predictions + sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2), 1, quantile, 0.975),
resError = sqrt(mean(model$model$sigma + model$model$tau) * model$sRe^2))
}
if (Bayes == FALSE & GAM == FALSE){
Est <- predict(model$model$gam,
data.frame(XPred,
Date2 = (time - mean(data$Date)) /
sd(data$Date)), se.fit = TRUE, type = "response")
if(estType == "1 SE"){
Est$fit <- Est$se.fit
}
if(estType == "2 SE"){
Est$fit <- Est$se.fit * 2
}
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
varM = Est$fit * (1-Est$fit)
} else {
varM = var(residuals(model$model$gam))
}
if(estType == "1 SETOTAL"){
Est$fit <- sqrt(Est$se.fit^2 + varM)
}
if(estType == "2 SETOTAL"){
Est$fit <- sqrt(Est$se.fit^2 + varM) * 2
}
if(estType == "1 SD Population"){
Est$fit <- sqrt(varM) * 1
}
if(estType == "2 SD Population"){
Est$fit <- sqrt(varM) * 2
}
if(estType == "Quantile"){
Est$fit <- Est$fit + qnorm(estQuantile) * Est$se.fit
}
if(estType == "QuantileTOTAL"){
Est$fit <- Est$fit + qnorm(estQuantile) *
sqrt(Est$se.fit^2 + varM)
}
XPred <- data.frame(XPred,
Est = Est$fit, Sd = Est$se.fit,
SDPop = sqrt(varM),
SdTotal = sqrt(Est$se.fit^2 + varM),
IntLower = Est$fit - 1.96 * Est$se.fit,
IntUpper = Est$fit + 1.96 * Est$se.fit,
IntLowerTotal = Est$fit - 1.96 *
sqrt(Est$se.fit^2 + varM),
IntUpperTotal = Est$fit + 1.96 *
sqrt(Est$se.fit^2 + varM),
resError = sqrt(mean((predict(model$model$gam, type = "response") - model$model$gam$y)^2)))
}
if(GAM == TRUE){
Predictions <- sapply(1:length(model$model),
function(x) predict(model$model[[x]],
x = cbind(XPred$Longitude, XPred$Latitude,
Date2 = (time - mean(data$Date)) / sd(data$Date))))
if(estType == "Mean"){
Est <- rowMeans(Predictions)
}
if(estType == "1 SE"){
Est <- apply(Predictions, 1, sd)
}
if(estType == "2 SE"){
Est <- apply(Predictions, 1, sd) * 2
}
if(estType == "Quantile"){
Est <- apply(Predictions, 1, quantile, estQuantile)
}
XPred <- data.frame(XPred,
Est = Est,
Sd = apply(Predictions, 1, sd),
IntLower = apply(Predictions, 1, quantile, 0.025),
IntUpper = apply(Predictions, 1, quantile, 0.975))
}
if (estType != "1 SE" && estType != "1 SETOTAL" && estType != "2 SE" &&
estType != "2 SETOTAL" &&
!is.null(limitz) && !missing(limitz)){
if (limitz == "0-1"){
XPred$Est <- pmin(1, XPred$Est)
XPred$Est <- pmax(0, XPred$Est)
}
if (limitz == "0-100"){
XPred$Est <- pmin(100, XPred$Est)
XPred$Est <- pmax(0, XPred$Est)
}
}
if (!is.null(dataCenter)){
# this is batch mode
XPredCenter <- lapply(1:nrow(dataCenter), function(x){
XPred %>%
extractXPredCenter(centerX = dataCenter[x, 1],
centerY = dataCenter[x, 2],
Radius = RadiusBatch,
batch = TRUE,
isThreeD = TRUE,
data = data,
time = dataCenter[x, 3])
})
centerEstimates <- lapply(1:nrow(dataCenter), function(x){
XPredCenter[[x]] %>%
extractCenterEstimates(batch = FALSE) # why is sd calculated here differently than in plotMap batch??
})
meanCenter <- sapply(1:nrow(dataCenter), function(x) centerEstimates[[x]]$mean)
sdCenter <- sapply(1:nrow(dataCenter), function(x) centerEstimates[[x]]$sd)
IntLower <- sapply(1:nrow(dataCenter),
function(x) signif(min(XPredCenter[[x]]$IntLower), 5))
IntUpper <- sapply(1:nrow(dataCenter),
function(x) signif(max(XPredCenter[[x]]$IntUpper), 5))
if(is.null(XPredCenter[[1]]$SDPop)) {
dataCenter <- cbind(dataCenter, time = time, mean = meanCenter, sd = sdCenter,
IntLower = IntLower, IntUpper = IntUpper)
} else {
SDPop <- sapply(1:nrow(dataCenter),
function(x) signif(sqrt(mean(XPredCenter[[x]]$SDPop, na.rm = TRUE)^2), 5))
sdCenterTotal <- sapply(1:nrow(dataCenter),
function(x) signif(sqrt(sum(c(sd(XPredCenter[[x]]$Est)^2,
mean(XPredCenter[[x]]$Sd^2),
mean(XPredCenter[[x]]$SdTotal)^2), na.rm = TRUE)), 5))
IntLowerTotal <- sapply(1:nrow(dataCenter),
function(x) signif(min(XPredCenter[[x]]$IntLowerTotal), 5))
IntUpperTotal <- sapply(1:nrow(dataCenter),
function(x) signif(max(XPredCenter[[x]]$IntUpperTotal), 5))
dataCenter <- cbind(dataCenter, time = time, mean = meanCenter, sd = sdCenter,
SDPop = SDPop,
sdTotal = sdCenterTotal,
IntLower = IntLower, IntUpper = IntUpper,
IntLowerTotal = IntLowerTotal, IntUpperTotal = IntUpperTotal)
}
return(dataCenter)
}
# keep $Est for later calculation of mean and sd for center
XPred$EstForCenter <- XPred$Est
if (interior > 0){
# this can remove all $Est
XPred$Est[draw == 0] <- NA
}
if (mask == TRUE){
XPred$Est[maskDraw == 0] <- NA
}
if(GAM == TRUE){
XPred$Est[is.na(XPred$Est) | XPred$Est < 0] <- 0
}
#if (!all(is.na(XPred$Est))){
levels <- pretty(c(rangez[1], rangez[2]), n = ncol)
if(GAM == TRUE){
colors <- c("#FFFFFF", colorRampPalette(brewer.pal(9, colors))(length(levels) - 2))
} else{
colors <- colorRampPalette(brewer.pal(9, colors))(length(levels) - 1)
}
if(reverseColors){
colors <- rev(colors)
}
levelsLegend <- levels
if(length(levels) > 25){
par(fg = NA, col="black")
levelsLegend <- pretty(c(rangez[1], rangez[2]),
n = pmin(20, ceiling(ncol / 2)))
}
cex4 <- 1
if(!all(is.na(XPred$Est)) &&
(max(abs(XPred$Est), na.rm = TRUE) > 9999 | max(abs(XPred$Est), na.rm = TRUE) < 0.05)){
cex4 <- 0.7
}
if(centerMap != "Europe"){
longitudes <- longitudesPac[order(longitudesPac)]
latitudes <- latitudes[order(latitudes, longitudesPac)]
XPredPlot <- data.frame(XPredPac[order(XPredPac$Latitude, XPredPac$Longitude),],
Est = XPred$Est[order(XPredPac$Latitude, XPredPac$Longitude)])
rangex <- rangexEU
} else {
XPredPlot <- XPred
}
if(!showModel){
XPredPlot$Est <- rep(NA, length(XPredPlot$Est))
}
if(titleMain){
main = ""
} else {
if(setAxisLabels){
main = mainLabel
} else {
main = independent
}
}
if(titleScale){
mainS = ""
} else {
if(setAxisLabels){
mainS = scLabel
} else {
mainS = paste("time:", time)
}
}
if(setAxisLabels){
xlab = xLabel
ylab = yLabel
} else {
xlab = "Longitude"
ylab = "Latitude"
}
filled.contour2(longitudes, latitudes, z = matrix(XPredPlot$Est, ncol = resolution),
contourType = contourType,
xlim = rangex, ylim = rangey, levels = levels,
col = colors,
showScale = showScale,
cex.axis = 1.5, cex.main = 1.5, cex.lab = 1.5,
asp = 1, xaxp = c(0, 100, 10), yaxp = c(0, 100, 10),
key.axes = axis(side = 4, at = levelsLegend, cex.axis = cex4),
key.title = title(main = main, cex.main = 0.8),
plot.title = {title(cex.lab = AxisSize, xlab = xlab, ylab = ylab, main = mainS)},
plot.axes = {
par(fg = "black", col="black");
if (terrestrial == 1){
if(centerMap != "Europe"){
sp::plot(Maps$ocean160, add = T, col = "lightblue", lwd = 1, border = NA);
sp::plot(Maps$ocean200, add = T, col = "lightblue", lwd = 1, border = NA);
} else {
sp::plot(Maps$ocean, add = T, col = "lightblue", lwd = 1);
}
}
if (terrestrial == -1){
if(centerMap != "Europe"){
sp::plot(Maps$land160, add = T, lwd = 1, col = "grey96", border = NA);
sp::plot(Maps$land200, add = T, lwd = 1, col = "grey96", border = NA);
} else {
sp::plot(Maps$land, add = T, lwd = 1, col = "grey96", border = NA);
}
}
if(centerMap != "Europe"){
sp::plot(Maps$coast160, add = T, lwd = 1);
sp::plot(Maps$coast200, add = T, lwd = 1);
} else {
sp::plot(Maps$coast, add = T, lwd = 1);
}
if (grid == TRUE){
if(centerMap != "Europe"){
sp::plot(Maps$grids160, add = T, col = "grey", lty = 2, xlim = c(0, 1));
sp::plot(Maps$grids200, add = T, col = "grey", lty = 2, xlim = c(0, 1));
} else {
sp::plot(Maps$grids, add = T, col = "grey", lty = 2, xlim = c(0, 1));
}
}
if(centerMap != "Europe"){
sp::plot(Maps$borders160, add = T, col = "darkgrey", lwd = 1);
sp::plot(Maps$borders200, add = T, col = "darkgrey", lwd = 1);
} else {
sp::plot(Maps$borders, add = T, col = "darkgrey", lwd = 1);
}
# add points and/or centroids ----
if (points == TRUE){
if(!is.null(pointLabels)){
pointLabels <- as.numeric(pointLabels)
pointLabels <- pointLabels - min(pointLabels, na.rm = TRUE)
pointLabels <- pointLabels / max(pointLabels, na.rm = TRUE)
pointSize <- (pointSize * pointLabels + 0.1) * 2
}
if(!is.null(pointColLabels)){
pointColLabels <- as.numeric(pointColLabels)
pointColLabels <- pointColLabels - min(pointColLabels, na.rm = TRUE)
pointColLabels <- pointColLabels / max(pointColLabels, na.rm = TRUE)
pColor <- (colorRampPalette(brewer.pal(9, colorsP))(101))[round(pointColLabels, 2) * 100]
}
if(cluster & !is.null(data$spatial_cluster)){
data <- data %>%
selectClusterGrouping(cluster, clusterResults)
# get centroids
if(clusterResults == 1){
data_names <- c("spatial_cluster", "long_centroid_spatial_cluster", "lat_centroid_spatial_cluster", "cluster")
centroids <- unique(data[, data_names])
centroids <- na.omit(centroids)
} else {
data_names <- c("temporal_group", "long_temporal_group_reference_point", "lat_temporal_group_reference_point", "cluster")
centroids <- unique(data[, data_names])
}
centroids <- centroids[order(centroids[,1]), ]
if(centerMap != "Europe"){
centroidsPac <- centroids
centroidsPac[,2][centroids[,2] < -20] <- centroidsPac[,2][centroids[,2] < -20] + 200
centroidsPac[,2][centroids[,2] >= -20] <- (- 160 + centroidsPac[,2][centroids[,2] >= -20])
centroids <- centroidsPac
}
}
# add points (location marks) ----
## if there is no clustering, or
## if not show only centroids
if (!cluster || clusterAll != "-1") {
dataPlot <- data %>%
centerPlotData(centerMap = centerMap)
## if there is no clustering, or
## if not "Show points for all times"
if (!cluster || clusterAll == "1") {
dataPlot <- dataPlot %>%
filterT(addU = addU, time = time)
}
dataPlot <- dataPlot %>%
selectClusterGrouping(cluster, clusterResults)
points(dataPlot$Latitude ~ dataPlot$Longitude,
col = getPColor(dataPlot, cluster, palName = clusterCol, pColor = pColor),
lwd = 2,
pch = pointShape,
cex = pointSize)
}
# add centroids ----
if (cluster & !is.null(data$spatial_cluster)) {
points(centroids[, 2:3],
col = getPColor(centroids, cluster, palName = clusterCol, pColor = pColor),
lwd = 2,
pch = pointShape,
cex = pointSize * 2.5)
text(centroids[, 2:3],
labels = sprintf("%s_%s",
ifelse(clusterResults == "0", "Group", "Cluster"),
centroids$cluster),
pos = 4,
cex = fontSize * 1.5,
col = fontCol,
family = fontType)
}
}
# add custom labels ----
if (!is.null(textLabels)) {
if (centerMap != "Europe") {
text(dataPac$Latitude ~ dataPac$Longitude,
labels = as.character(textLabels), pos = 4,
cex = fontSize, col = fontCol, family = fontType)
} else {
text(data$Latitude ~ data$Longitude,
labels = as.character(textLabels), pos = 4,
cex = fontSize, col = fontCol, family = fontType)
}
}
# add custom points ----
if (!is.null(pointDat) & NROW(pointDat) > 0) {
points(x = pointDat$x, y = pointDat$y, cex = pointDat$pointSize, col = pointDat$pointColor, pch = 16)
text(pointDat$y ~ pointDat$x, labels = pointDat$label, pos = 4, cex = 1.75)
}
# update axis if center Pacific
if (centerMap != "Europe") {
lab <- pretty(rangex)
lab[pretty(rangex) >= 20] <- lab[pretty(rangex) >= 20] - 200
lab[pretty(rangex) < 20] <- lab[pretty(rangex) < 20] + 160
axis(1, at = pretty(rangex), labels = lab, cex.axis = AxisLSize);
} else{
axis(1, cex.axis = AxisLSize);
}
axis(2, cex.axis = AxisLSize);
}
)
# } else {
# plot(1:10, xlim = range(data$Longitude), ylim = range(data$Latitude),
# xlab = "Longitude", ylab = "Latitude")
# plot(Maps$ocean, add = T, col = "lightblue", lwd = 1)
# sp::plot(Maps$grids, add = T, col = "grey", lty = 2, xlim = c(0, 1))
# sp::plot(Maps$borders, add = T, col = "darkgrey", lwd = 1)
# text(mean(data$Longitude), mean(data$Latitude), cex = 1,
# "No estimates to plot, please adjust
# \"Display up to max standard error\" slider", col = "red")
# }
if (arrow == TRUE){
if(showScale == TRUE){
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
} else {
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
}
}
if (scale == TRUE){
if(showScale == TRUE){
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
} else {
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
}
}
if(plotRetNull){
return(NULL)
} else {
return(list(XPred = XPred))
}
}
#' Plots difference or similarity map
#'
#' @param XPred return object of similarityMap or createDifferenceMap functions
#' @param estType one of "Mean", "SE" and "Quantile". defaults to "Mean"
#' @param estQuantile quantile (only applicable if estType = "Quantile")
#' @param type one of "similarity" or "difference" determining the type of map
#' @param independent name of independent variable shown in plot
#' @param arrow display north arrow TRUE/FALSE
#' @param scale display scale TRUE/FALSE
#' @param terrestrial show only estimates on land masses (1), oceans (-1) or all (0)
#' @param grid show coordinate grid TRUE/FALSE
#' @param ncol number of colors for estimates
#' @param colors color scheme of estimates from RColorBrewer. defaults to "RdYlGn"
#' @param reverseColors inverse colour scheme
#' @param rangex range of longitude values (x axis limits)
#' @param rangey range of latitude values (y axis limits)
#' @param rangez range of estimated values (z axis limits)
#' @param centerMap center of map, one of "Europe" and "Pacific"
#' @param showValues boolean show values in plot?
#' @param simValues if showValues: list of simulated values
#' @param dataCenter data for batch estimation
#' @param RadiusBatch radius
#' @param showModel show model
#' @param titleMain show main title
#' @param titleScale show scale title
#' @param setAxisLabels show axis/main/scale labels
#' @param mainLabel main label
#' @param yLabel y lab label
#' @param xLabel y lab label
#' @param scLabel scale lab labels
#' @param northSize size of north arrow
#' @param scalSize size of scale
#' @param scaleX scale x orientation
#' @param scaleY scale y orientation
#' @param NorthX north arrow x orientation
#' @param NorthY north arrow y orientation
#' @param AxisSize axis title font size
#' @param AxisLSize axis label font size
#' @param pointDat data frame of points to add to plot
#' @inheritParams filled.contour2
#'
#' @export
plotDS <- function(XPred,
estType = "mean",
estQuantile = 0.95,
type = "similarity",
independent = "",
rangex = range(XPred$Longitude, na.rm = TRUE),
rangey = range(XPred$Latitude, na.rm = TRUE),
rangez = range(XPred$Est, na.rm = TRUE),
contourType = c("filled.contour", "contour"),
showModel = TRUE,
showScale = TRUE,
ncol = 10, colors = "RdYlGn",
centerMap = "Europe",
reverseColors = FALSE, terrestrial = 1, grid = TRUE,
arrow = TRUE, scale = TRUE,
simValues = NULL,
showValues = FALSE,
dataCenter = NULL, RadiusBatch = NULL,
titleMain = TRUE,
titleScale = TRUE,
setAxisLabels = FALSE,
mainLabel = "",
yLabel = "Latitude",
xLabel = "Longitude",
scLabel = "",
northSize = 0.2,
scalSize = 0.1,
scaleX = 0,
scaleY = 0.1,
NorthX = 0.025,
NorthY = 0.925,
AxisSize = 1,
AxisLSize = 1,
pointDat = NULL){
options(scipen = 999)
contourType <- match.arg(contourType)
RadiusBatch <- RadiusBatch / 111
minRangeFactor <- 0.75
if ((diff(rangex) / diff(rangey)) < minRangeFactor) {
rangex[1] <- max(-180, mean(rangex) - minRangeFactor / 2 * diff(rangey))
rangex[2] <- min(180, mean(rangex) + minRangeFactor / 2 * diff(rangey))
}
minRangeFactor <- 0.5
if((diff(rangey) / diff(rangex)) < minRangeFactor){
rangey[1] <- max(-90, mean(rangey) - minRangeFactor / 2 * diff(rangex))
rangey[2] <- min(90, mean(rangey) + minRangeFactor / 2 * diff(rangex))
}
if (!is.null(dataCenter)){
# this is batch mode
XPred2 <- do.call("rbind", lapply(1:nrow(dataCenter), function(x) {
longitudes <- seq(dataCenter[x, 1] - RadiusBatch,
dataCenter[x, 1] + RadiusBatch, length.out = 4)
latitudes <- seq(dataCenter[x, 2] - RadiusBatch,
dataCenter[x, 2] + RadiusBatch, length.out = 4)
XPred2 <- expand.grid(Longitude = longitudes, Latitude = latitudes)
cbind(XPred2, id = x)
}))
closestMean <- sapply(1:nrow(XPred2), function(x)
XPred$Est[which.min((XPred2$Longitude[x] - XPred$Longitude) ^ 2 +
(XPred2$Latitude[x] - XPred$Latitude) ^ 2
)])
closestSD <- sapply(1:nrow(XPred2), function(x)
XPred$Sd[which.min((XPred2$Longitude[x] - XPred$Longitude) ^ 2 +
(XPred2$Latitude[x] - XPred$Latitude) ^ 2
)])
closestDist <- sapply(1:nrow(XPred2), function(x)
min((XPred2$Longitude[x] - XPred$Longitude) ^ 2 +
(XPred2$Latitude[x] - XPred$Latitude) ^ 2
))
closestMean[closestDist > RadiusBatch / 5] <- NA
closestSD[closestDist > RadiusBatch / 5] <- NA
XPred2$Mean <- closestMean
XPred2$SD <- closestSD
XPred2 <- do.call("rbind", lapply(1:max(XPred2$id), function(x) {
tmp <- XPred2[XPred2$id == x,]
data.frame(
Longitude = tmp$Longitude[1],
Latitude = tmp$Latitude[1],
mean = signif(mean(tmp$Mean, na.rm = TRUE), 3),
sd = signif(mean(tmp$SD, na.rm = TRUE) +
sd(tmp$Mean, na.rm = TRUE), 3)
)
})
)
XPred2$IntLower <- XPred2$mean - 1.96 * XPred2$sd
XPred2$IntUpper <- XPred2$mean + 1.96 * XPred2$sd
return(XPred2)
}
if(estType == "1 SE"){
XPred$Est <- XPred$Sd
}
if(estType == "2 SE"){
XPred$Est <- 2 * XPred$Sd
}
if(estType == "Significance (p-value)"){
XPred$Est <- pnorm(-abs(XPred$Est), 0, XPred$Sd) * 2
}
if(estType == "Significance (z-value)"){
XPred$Est <- XPred$Est / XPred$Sd
}
if(estType == "Quantile"){
XPred$Est <- XPred$Est + qnorm(estQuantile) * XPred$Sd
if(type == "similarity"){
XPred$Est <- pmax(0, XPred$Est)
}
}
#if (!all(is.na(XPred$Est))){
if(type == "similarity"){
z <- matrix(XPred$Est, ncol = length(unique(XPred$Latitude)))
}
if(type == "difference"){
z <- matrix(XPred$Est, ncol = length(unique(XPred$Latitude)))
}
Maps <- loadMaps()
levels <- pretty(c(rangez[1], rangez[2]), n = ncol)
colors <- colorRampPalette(brewer.pal(9, colors))(length(levels) - 1)
if(reverseColors){
colors <- rev(colors)
}
levelsLegend <- levels
if(length(levels) > 25){
par(fg = NA, col="black")
levelsLegend <- pretty(c(rangez[1], rangez[2]),
n = pmin(20, ceiling(ncol / 2)))
}
# keep $Est for later calculation of mean and sd for center
XPred$EstForCenter <- XPred$Est
if(centerMap != "Europe"){
XPredPac <- XPred
XPredPac$Longitude[XPred$Longitude < -20] <- XPredPac$Longitude[XPred$Longitude < -20] + 200
XPredPac$Longitude[XPred$Longitude >= -20] <- (- 160 + XPredPac$Longitude[XPred$Longitude >= -20])
XPredPlot <- data.frame(XPredPac[order(XPredPac$Latitude, XPredPac$Longitude),])
z <- matrix(XPredPlot$Est, ncol = length(unique(XPredPlot$Latitude)))
} else {
XPredPlot <- XPred
}
if(!showModel){
z <- matrix(NA, ncol = length(unique(XPredPlot$Latitude)), nrow = length(unique(XPredPlot$Longitude)))
}
cex4 <- 1
if(max(abs(z), na.rm = TRUE) > 9999 | max(abs(z), na.rm = TRUE) < 0.05){
cex4 <- 0.7
}
if(titleMain){
main = ""
} else {
if(setAxisLabels){
main = mainLabel
} else {
main = independent
}
}
if(titleScale){
mainS = ""
} else {
if(setAxisLabels){
mainS = scLabel
} else {
mainS = independent
}
}
if(setAxisLabels){
xlab = xLabel
ylab = yLabel
} else {
xlab = "Longitude"
ylab = "Latitude"
}
if (any(unique(XPredPlot$Longitude) != sort(unique(XPredPlot$Longitude))) |
any(unique(XPredPlot$Latitude) != sort(unique(XPredPlot$Latitude)))){
return("The meridian should match the corresponding region. Spherical coordinates should also match. Try to adjust the map centering option.")
}
filled.contour2(unique(XPredPlot$Longitude), unique(XPredPlot$Latitude),
z = z,
contourType = contourType,
xlim = rangex, ylim = rangey, levels = levels,
col = colors,
showScale = showScale,
cex.axis = 1.5, cex.main = 1.5, cex.lab = 1.5,
asp = 1, key.axes = axis(side = 4, at = levelsLegend, cex.axis = cex4),
key.title = title(main = main, cex.main = 0.8),
plot.title = {title(cex.lab = AxisSize, xlab = xlab, ylab = ylab, main = mainS)},
plot.axes = {
par(fg = "black", col="black");
if (terrestrial == 1){
if(centerMap != "Europe"){
sp::plot(Maps$ocean160, add = T, col = "lightblue", lwd = 1, border = NA);
sp::plot(Maps$ocean200, add = T, col = "lightblue", lwd = 1, border = NA);
} else {
sp::plot(Maps$ocean, add = T, col = "lightblue", lwd = 1);
}
}
if (terrestrial == -1){
if(centerMap != "Europe"){
sp::plot(Maps$land160, add = T, lwd = 1, col = "grey96", border = NA);
sp::plot(Maps$land200, add = T, lwd = 1, col = "grey96", border = NA);
} else {
sp::plot(Maps$land, add = T, lwd = 1, col = "grey96", border = NA);
}
}
if(centerMap != "Europe"){
sp::plot(Maps$coast160, add = T, lwd = 1);
sp::plot(Maps$coast200, add = T, lwd = 1);
} else {
sp::plot(Maps$coast, add = T, lwd = 1);
}
if (grid == TRUE){
if(centerMap != "Europe"){
sp::plot(Maps$grids160, add = T, col = "grey", lty = 2, xlim = c(0, 1));
sp::plot(Maps$grids200, add = T, col = "grey", lty = 2, xlim = c(0, 1));
} else {
sp::plot(Maps$grids, add = T, col = "grey", lty = 2, xlim = c(0, 1));
}
}
if(centerMap != "Europe"){
sp::plot(Maps$borders160, add = T, col = "darkgrey", lwd = 1);
sp::plot(Maps$borders200, add = T, col = "darkgrey", lwd = 1);
} else {
sp::plot(Maps$borders, add = T, col = "darkgrey", lwd = 1);
}
if(centerMap != "Europe"){
lab <- pretty(rangex)
lab[pretty(rangex) >= 20] <- lab[pretty(rangex) >= 20] - 200
lab[pretty(rangex) < 20] <- lab[pretty(rangex) < 20] + 160
axis(1, at = pretty(rangex), labels = lab, cex.axis = AxisLSize);
} else{
axis(1, cex.axis = AxisLSize);
}
axis(2, cex.axis = AxisLSize);
if(!is.null(pointDat) & NROW(pointDat) > 0){
points(x = pointDat$x, y = pointDat$y, cex = pointDat$pointSize, col = pointDat$pointColor, pch = 16)
text(pointDat$y ~ pointDat$x, labels = pointDat$label, pos = 4, cex = 1.75)
}
if (showValues == TRUE) {
op <- par(family = "mono")
titel = paste(capture.output(do.call("rbind", simValues)), collapse = '\n')
legend("bottomright", legend = c(""), pch = NA, title = titel,
text.width = strwidth(titel)[1]/2, bty = "n", cex = 2, text.font = 2)
par(op)
}
}
)
# } else {
# plot(1:10, xlim = range(XPred$Longitude), ylim = range(XPred$Latitude),
# xlab = "Longitude", ylab = "Latitude")
# plot(Maps$ocean, add = T, col = "lightblue", lwd = 1)
# sp::plot(Maps$grids, add = T, col = "grey", lty = 2, xlim = c(0, 1))
# sp::plot(Maps$borders, add = T, col = "darkgrey", lwd = 1)
# text(mean(XPred$Longitude), mean(XPred$Latitude), cex = 1,
# "No estimates to plot, please adjust \"Display up to max standard error\" slider",
# col = "red")
# }
if (arrow == TRUE){
if(showScale == TRUE){
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
} else {
north.arrow(rangex[1] + diff(rangey) * NorthX, rangey[1] + diff(rangey) * NorthY,
diff(rangey) * 0.04 * northSize * 5,
1 * ((2 * diff(rangey)) / diff(rangex)) ^ 0.3* northSize * 5, c(0.5, - 0.25))
}
}
if (scale == TRUE){
if(showScale == TRUE){
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
} else {
maps::map.scale(x = rangex[1] + diff(rangex) * scaleX,
y = rangey[1] + diff(rangey) * scaleY,
rel = scalSize * ((2 * diff(rangey)) / diff(rangex)) ^ 0.15,
cex = 0.75 * ((2 * diff(rangey)) / diff(rangex)) ^ -0.2, ratio = FALSE)
}
}
return(list(XPred = XPred))
}
#' Filled Countour 2
#' Wrapper for contour plot
#'
#' @param x x values
#' @param y y values
#' @param z z values
#' @param contourType one of "filled.contour" or "contour"
#' @param xlim x limits
#' @param ylim y limits
#' @param zlim z limits
#' @param levels levels
#' @param nlevels number of levels
#' @param showScale show colour scale
#' @param color.palette color palette
#' @param col colors
#' @param plot.title plot title
#' @param plot.axes plot axes
#' @param key.title key title
#' @param key.axes key axes
#' @param asp aspect ratio
#' @param xaxs x axis style
#' @param yaxs y axis style
#' @param las label style
#' @param axes show axes
#' @param frame.plot frame plot
#' @param ... additional arguments
filled.contour2 <- function(x = seq(0, 1, length.out = nrow(z)),
y = seq(0, 1, length.out = ncol(z)),
z,
contourType = c("filled.contour", "contour"),
xlim = range(x, finite = TRUE),
ylim = range(y, finite = TRUE), zlim = range(z, finite = TRUE),
levels = pretty(zlim, nlevels), nlevels = 20, showScale = TRUE,
color.palette = cm.colors, col = color.palette(length(levels) - 1),
plot.title, plot.axes, key.title, key.axes, asp = NA, xaxs = "i",
yaxs = "i", las = 1, axes = TRUE, frame.plot = axes, ...)
{
contourType <- match.arg(contourType)
if (missing(z)) {
if (!missing(x)) {
if (is.list(x)) {
z <- x$z
y <- x$y
x <- x$x
}
else {
z <- x
x <- seq.int(0, 1, length.out = nrow(z))
}
}
else stop("no 'z' matrix specified")
}
else if (is.list(x)) {
y <- x$y
x <- x$x
}
if (any(diff(x) <= 0) || any(diff(y) <= 0))
stop("increasing 'x' and 'y' values expected")
# setup layout
mar.orig <- (par.orig <- par(c("mar", "las", "mfrow")))$mar
on.exit(par(par.orig))
par(las = las)
pin1 <- par("pin")
ratioLim <- abs(diff(ylim)) / abs(diff(xlim))
if (showScale && contourType == "filled.contour") {
# set up colour legend
w <- (3 + mar.orig[2L]) * par("csi") * 2.54
layout(matrix(c(2, 1), ncol = 2L), widths = c(1, lcm(w)))
mar <- mar.orig
mar[4L] <- mar[2L]
mar[2L] <- 1
par(mar = mar)
a = (pin1[1] + par("mai")[2] + par("mai")[4])
b = (pin1[2] + par("mai")[1] + par("mai")[3])
ratioXY <- (a / b)
if ((1 / ratioLim) >= ratioXY) {
par(plt = getPlotRegion(bottom = 0.5, ratioLim = ratioLim, ratioXY = ratioXY))
}
if ((1 / ratioLim) < ratioXY) {
par(plt = c(0.15, 0.5, 0.15, 0.9))
}
# create colour legend
plot.new()
plot.window(xlim = c(0, 1), ylim = range(levels), xaxs = "i",
yaxs = "i")
rect(0, levels[-length(levels)], 1, levels[-1L], col = col)
if (missing(key.axes)) {
if (axes)
if (max(abs(z), na.rm = TRUE) < 10000) {
axis(4)
} else {
axis(4, cex.axis = 0.7)
}
}
else key.axes
box()
if (!missing(key.title))
key.title
}
# set up (filled.)contour
mar <- mar.orig
mar[4L] <- 1
par(mar = mar)
a = (pin1[1] + par("mai")[2] + par("mai")[4])
b = (pin1[2] + par("mai")[1] + par("mai")[3])
ratioXY <- (a / b)
if ((1 / ratioLim) >= ratioXY) {
par(plt = getPlotRegion(bottom = 0.975, ratioLim = ratioLim, ratioXY = ratioXY))
}
if ((1 / ratioLim) < ratioXY) {
add <- 1 / ratioXY / 2 * 0.75 / ratioLim
par(plt = c(0.975 - 2 * add, 0.975, 0.15, 0.9))
}
# create (filled.)contour
plot.new()
if (contourType == "contour") {
# contour plot ----
contour(x = x, y = y, z = z, nlevels = length(levels), col = col,
xlim = xlim, ylim = ylim, xaxs = xaxs, yaxs = yaxs, asp = asp)
} else {
# filled.contour plot ----
plot.window(xlim, ylim, "", xaxs = xaxs, yaxs = yaxs, asp = asp)
.filled.contour(x, y, z, levels, col)
}
# set plot axis and background
if (missing(plot.axes)) {
if (axes) {
title(main = "", xlab = "", ylab = "")
Axis(x, side = 1)
Axis(y, side = 2)
}
}
else plot.axes
if (frame.plot)
box()
if (missing(plot.title))
title(...)
else plot.title
invisible()
}
getPlotRegion <- function(left = 0.1, bottom = 0.975, ratioLim = 1, ratioXY = 1) {
c(left, bottom, 0.525 - ratioLim * ratioXY / 2 * 0.875, 0.525 + ratioLim * ratioXY / 2 * 0.875)
}
north.arrow = function(x, y, h, c, adj) {
polygon(c(x, x, x + h/2), c(y - h, y, y - (1 + sqrt(3)/2) * h),
col = "black", border = NA)
polygon(c(x, x + h/2, x, x - h/2), c(y - h, y - (1 + sqrt(3)/2) *
h, y, y - (1 + sqrt(3)/2) * h))
text(x, y, "N", adj = adj, cex = c)
}
#' Plots time course map of a spatio-temporal model from estimateMap3D() function
#'
#' @param model return object of a spatio-temporal model from estimateMap3D() function
#' @param IndSelect for categorical model: selected category
#' @param independent name of independent variable shown in plot
#' @param trange range of longitude values (x axis limits)
#' @param resolution temporal grid resolution of displayed (higher is slower but better quality)
#' @param centerX longitude value to display time course plot for
#' @param centerY latitude value to display time course plot for
#' @param rangey 2-element vector of time interval to show
#' @param seType setype
#' @param pointDat add points/lines to plot
#' @param pointsTime should nearby points be plotted?
#' @param returnPred should the prediction be returned?
#' @param intTime should uncertainty intervals of nearby points be plotted?
#' @param rangePointsTime range of nearby points in km
#' @param limitz z limit range
#' @param formatTimeCourse parameters for the plot format, e.g. axesDecPlace, nLabelsX, nLabelsY
#'
#' @export
plotTimeCourse <- function(model, IndSelect = NULL,
independent = "", trange = range(model$data$Date),
resolution = 500, centerX = NA,
centerY = NA, rangey = NULL,
seType = "2",
pointDat = NULL,
pointsTime = FALSE,
returnPred = FALSE,
intTime = FALSE,
rangePointsTime = 500,
limitz = NULL,
formatTimeCourse = NULL){
sdValue <- 1
if(as.numeric(seType) > 5){
sdValue <- 2
}
minVal <- -Inf
maxVal <- Inf
if(!is.null(limitz) && limitz == "0-100"){
minVal <- 0
maxVal <- 100
}
if(!is.null(limitz) && limitz == "0-1"){
minVal <- 0
maxVal <- 1
}
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
if(IndSelect == "" | is.null(IndSelect)){
return(NULL)
}
model$model <- model$model[[IndSelect]]
minVal <- max(minVal, 0)
maxVal <- min(maxVal, 1)
}
Bayes = TRUE
GAM = FALSE
if("gamm" %in% class(model$model)){
Bayes = FALSE
}
if("kde" %in% class(model$model)){
GAM = TRUE
}
data <- model$data
if (is.null(data)) return(NULL)
if (is.na(centerX)){centerX <- signif(mean(data$Longitude, na.rm = TRUE), 3)}
if (is.na(centerY)){centerY <- signif(mean(data$Latitude, na.rm = TRUE), 3)}
time <- seq(trange[1], trange[2], length.out = resolution)
XPred <- data.frame(time,
Date2 = (time - mean(data$Date)) /
sd(data$Date),
Longitude2 = (centerX - mean(data$Longitude)) / sd(data$Longitude),
Latitude2 = (centerY - mean(data$Latitude)) / sd(data$Latitude),
Longitude = centerX,
Latitude = centerY)
if (Bayes == TRUE & GAM == FALSE){
sc <- model$sc
PredMatr <- Predict.matrix(sc, data = XPred)
betas <- model$model$beta
betaSigma <- model$model$betaSigma
Predictions <-
sapply(1:nrow(betas), function(x)
PredMatr %*% betas[x, ] * model$sRe + model$mRe)
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
Predictions <- invLogit(Predictions)
}
if(!is.null(betaSigma)){
PredMatrV <- Predict.matrix(model$scV, data = XPred)
PredictionsSigma <-
rowMeans(sqrt(sapply(1:nrow(betaSigma), function(x)
exp((PredMatrV %*% betaSigma[x, ])) / model$model$sigma[x]) * model$sRe^2))
} else {
if(!is.null(model$IndependentType) && model$IndependentType != "numeric"){
PredictionsSigma <- sqrt(Predictions * (1-Predictions))
} else {
PredictionsSigma <- sqrt(mean(model$model$sigma) * model$sRe^2)
}
}
if(!is.null(betaSigma)){
SdTotal <- sqrt((PredictionsSigma)^2 + apply(Predictions, 1, sd)^2)
} else {
SdTotal <- sqrt(PredictionsSigma^2 + apply(Predictions, 1, sd)^2)
}
XPred <-
data.frame(XPred,
Est = rowMeans(Predictions),
Sd = apply(Predictions, 1, sd),
SdTotal = SdTotal,
PredictionsSigma = PredictionsSigma,
IntLower = pmax(minVal, pmin(maxVal, rowMeans(Predictions) - sdValue * apply(Predictions, 1, sd))),
IntUpper = pmax(minVal, pmin(maxVal, rowMeans(Predictions) + sdValue * apply(Predictions, 1, sd))))
mainlab <- paste0("Estimate of ", independent, " in time course at coordinates ",
"(", centerY,",", centerX,")" ," with credible intervals")
}
if (Bayes == FALSE & GAM == FALSE){
Est <- predict(model$model$gam, XPred, se.fit = TRUE, type = "response")
XPred <- data.frame(XPred, Est = Est$fit, Sd = Est$se.fit,
SdTotal = sqrt(Est$se.fit^2 + mean(residuals(model$model$gam)^2)),
PredictionsSigma = sd(residuals(model$model$gam)),
IntLower = pmax(minVal, pmin(maxVal, Est$fit - sdValue * Est$se.fit)),
IntUpper = pmax(minVal, pmin(maxVal, Est$fit + sdValue * Est$se.fit))
# IntLowerTotal = Est$fit - 1.96 * sqrt(Est$se.fit^2 + mean(residuals(model$model$gam)^2)),
# IntUpperTotal = Est$fit + 1.96 * sqrt(Est$se.fit^2 + mean(residuals(model$model$gam)^2))
)
mainlab <- paste0("Estimate of ", independent, " in time course at coordinates ",
"(", centerY,",", centerX,")" ," with confidence intervals")
}
if (GAM == TRUE){
EstTemp <- sapply(1:length(model$model), function(x) predict(model$model[[x]],
x = cbind(XPred$Longitude, XPred$Latitude, XPred$Date2)))
estQuantile <-
XPred <- data.frame(XPred,
Est = rowMeans(EstTemp),
Sd = apply(EstTemp, 1, sd),
IntLower = pmax(0, apply(EstTemp, 1, quantile, 1 - pnorm(sdValue))),
IntUpper = pmax(0, apply(EstTemp, 1, quantile, pnorm(sdValue))))
mainlab <- paste0("Density estimate in time course at coordinates ",
"(", centerY,",", centerX,").")
}
ylims <- c(min(XPred$Est - sdValue * XPred$Sd), max(XPred$Est + sdValue * XPred$Sd))
if(!is.null(rangey)){
ylims[!is.na(rangey)] = rangey[!is.na(rangey)]
}
plot(
pmax(minVal, pmin(maxVal, XPred$Est)) ~ time,
type = "l",
ylim = ylims,
xlim = c(min(time), max(time)),
ylab = independent,
xlab = "time",
main = mainlab,
xaxt = 'n',
yaxt = 'n'
)
if (!is.null(formatTimeCourse)) {
addFormattedAxis(
axis = "x",
min = min(time),
max = max(time),
nLabels = formatTimeCourse$nLabelsX,
decPlace = formatTimeCourse$axesDecPlace
)
addFormattedAxis(
axis = "y",
min = ylims[1],
max = ylims[2],
nLabels = formatTimeCourse$nLabelsY,
decPlace = formatTimeCourse$axesDecPlace
)
}
if(seType %in% c("2", "4", "6", "9")){
polygon(c(rev(time), time), pmax(minVal, pmin(maxVal, c(rev(XPred$IntUpper), XPred$IntLower))),
col = 'grey90', border = NA)
lines(pmax(minVal, pmin(maxVal, (XPred$Est))) ~ (time), lty = 1)
lines(pmax(minVal, pmin(maxVal, (XPred$IntUpper))) ~ (time), lty = 2)
lines(pmax(minVal, pmin(maxVal, (XPred$IntLower))) ~ (time), lty = 2)
}
if(seType %in% c("5", "7")){
polygon(c(rev(time), time), pmax(minVal, pmin(maxVal,
c(rev(XPred$Est - sdValue * XPred$PredictionsSigma),
XPred$Est + sdValue * XPred$PredictionsSigma))),
col = 'grey90', border = NA)
lines(pmax(minVal, pmin(maxVal, (XPred$Est))) ~ (time), lty = 1)
lines(pmax(minVal, pmin(maxVal, (XPred$Est - sdValue * XPred$PredictionsSigma))) ~ (time), lty = 2)
lines(pmax(minVal, pmin(maxVal, (XPred$Est + sdValue * XPred$PredictionsSigma))) ~ (time), lty = 2)
}
if(seType %in% c("3", "4", "8", "9")){
lines(pmax(minVal, pmin(maxVal,(XPred$Est + sdValue * XPred$SdTotal))) ~ (time), lty = 3)
lines(pmax(minVal, pmin(maxVal,(XPred$Est - sdValue * XPred$SdTotal))) ~ (time), lty = 3)
}
if(!is.null(pointDat) & NROW(pointDat) > 0){
for(i in 1:nrow(pointDat)){
if(!is.na(pointDat$y[i]) & !is.na(pointDat$x[i])){
points(pointDat$y[i] ~ pointDat$x[i], col = pointDat$pointColor[i],
cex = pointDat$pointSize[i], pch = 19)
if(!is.na(pointDat$ymin[i]) & !is.na(pointDat$ymax[i])){
lines(x = rep(pointDat$x[i], 2), y = c(pointDat$ymin[i], pointDat$ymax[i]),
lwd = pointDat$pointSize[i])
}
if(!is.na(pointDat$xmin[i]) & !is.na(pointDat$xmax[i])){
lines(y = rep(pointDat$y[i], 2), x = c(pointDat$xmin[i], pointDat$xmax[i]),
lwd = pointDat$pointSize[i])
}
}
}
}
if(returnPred){
return(XPred[, !(names(XPred) %in% c("Date2", "Longitude2", "Latitude2"))])
}
if(pointsTime){
pointPlotData <- model$data[(sqrt((model$data$Longitude - centerX)^2 +
(model$data$Latitude - centerY)^2) * 111) < rangePointsTime, ]
if(NROW(pointPlotData) > 0){
if(model$independent == ""){
pointPlotData$ind <- 0
ind <- "ind"
} else {
if(!is.null(model$IndependentType) && model$IndependentType == "numeric"){
ind <- model$independent
} else {
if(IndSelect == "" | is.null(IndSelect)){
ind <- model$independent
} else {
ind <- IndSelect
}
}
}
points(pointPlotData[, ind] ~ pointPlotData$Date)
if(intTime){
for(i in 1:nrow(pointPlotData)){
lines(c((pointPlotData$Date[i] - 2 * pointPlotData$Uncertainty[i]),
(pointPlotData$Date[i] + 2 * pointPlotData$Uncertainty[i])),
c(pointPlotData[, ind][i], pointPlotData[, ind][i]))
}
pointPlotData <- pointPlotData[(pointPlotData$Date + 2 * pointPlotData$Uncertainty > trange[1]) &
(pointPlotData$Date - 2 * pointPlotData$Uncertainty < trange[2]), ]
} else {
pointPlotData <- pointPlotData[(pointPlotData$Date > trange[1]) &
(pointPlotData$Date < trange[2]), ]
}
}
return(pointPlotData)
}
return(NULL)
}
#' Add Formatted Axis
#'
#' @param axis (character) axis to add, either "x" or "y"
#' @param min (numeric) position of 1st label
#' @param max (numeric) position of last label
#' @param nLabels (numeric) number of displayed labels
#' @param decPlace (numeric) number of the label's decimal places
addFormattedAxis <- function(axis, min, max, nLabels = 7, decPlace = 0) {
labelPositions <- seq(min, max, length.out = nLabels)
axisType <- switch(axis,
"x" = 1,
"y" = 2)
axis(axisType,
at = labelPositions,
labels = sprintf(paste('%1.', decPlace, 'f', sep = ""), labelPositions))
return(NULL)
}
plotTimeIntervals <- function(Model,
trange = c(0, 1000),
AxisSize = 1,
AxisLSize = 1,
cluster = FALSE,
clusterCol = "Set1",
clusterResults = 0
){
dat <- Model$data
dat <- dat %>%
selectClusterGrouping(cluster = cluster,
clusterResults = clusterResults)
if(!is.null(dat$spatial_cluster)){
dat$cluster_color <- colorRampPalette(brewer.pal(8, clusterCol))(max(dat$cluster, na.rm=TRUE))[dat$cluster]
dat <- dat[!is.na(dat$cluster),]
dat$cluster <- factor(dat$cluster)
g <- ggplot(dat, aes_(~Date, ~cluster)) + theme_light() + coord_cartesian(xlim = trange) +
theme(panel.grid.major.x = element_blank(),
panel.grid.minor.x = element_blank(),
axis.text=element_text(size=12 * AxisLSize),
axis.title=element_text(size=14 * AxisSize), legend.position = "none") +
geom_point(color=dat$cluster_color, position = position_dodge(0.3), alpha = 0.3) +
geom_errorbar(
aes_(xmin = ~ Date-2*Uncertainty, xmax = ~ Date + 2*Uncertainty),
color=dat$cluster_color,
position = position_dodge(0.3), width = 0.1, alpha = 0.3) +
ylab(ifelse(clusterResults == "0", "temporal_group", "spatial_cluster"))
print(g)
} else {
plot(1, cex = 0.1)
text(x = 1, y= 1, "Please run KernelTimeR with Cluster Analysis (left Panel)")
}
}
roundUpNice <- function(x, nice=c(1,2,5,10)) {
y <- x
if(y < 0){
x <- -x
}
if(length(x) != 1) stop("'x' must be of length 1")
x <- 10^floor(log10(x)) * nice[[which(x <= 10^floor(log10(x)) * nice)[[1]]]]
if(y < 0){
x <- -x
}
x
}
createSimilarityMap <- function(XPredList, pointList,
includeUncertainty = TRUE,
normalize = FALSE,
normalType = "1",
weightProb = FALSE,
weightMap = NULL,
negZero = TRUE,
invWeight = FALSE,
weightDecay = 1){
XPredList <- lapply(1:length(XPredList), function(x){
similarityMap(XPredList[[x]], pointList[[x]], includeUncertainty)
})
combineSimilarityMaps(XPredList,
normalize = normalize,
normalType = normalType,
weightProb = weightProb,
weightMap = weightMap,
negZero = negZero,
invWeight = invWeight,
weightDecay = weightDecay)
}
similarityMap <- function(XPred, point, includeUncertainty = TRUE) {
if (includeUncertainty == TRUE) {
density <- lapply(1:nrow(point), function(x)
log(rowMeans(sapply(1:1000,
function(y)
dnorm(
rnorm(1, point$mean[x], point$sd[x]),
mean = XPred$Est,
sd = sqrt(XPred$Sd ^ 2 + XPred$resError ^ 2),
log = FALSE
)))))
densitySD <- lapply(1:nrow(point), function(x)
apply((sapply(1:1000,
function(y)
dnorm(
rnorm(1, point$mean[x], point$sd[x]),
mean = XPred$Est,
sd = sqrt(XPred$Sd ^ 2 + XPred$resError ^ 2),
log = FALSE
))), 1, sd))
XPred$densitySd <- Reduce("*", densitySD)
} else {
density <- lapply(1:nrow(point), function(x)
dnorm(
point$mean[x],
mean = XPred$Est,
sd = sqrt(XPred$Sd ^ 2 + XPred$resError ^ 2),
log = TRUE)
)
XPred$densitySd <- 0
}
XPred$density <- exp(Reduce("+", density))
XPred$Est <- XPred$density
XPred$Sd <- XPred$densitySd
XPred$IntLower <- pmax(0, XPred$Est - 1.96 * XPred$Sd)
XPred$IntUpper <- XPred$Est + 1.96 * XPred$Sd
return(XPred)
}
combineSimilarityMaps <- function(XPredList,
normalize = FALSE,
normalType = "1",
weightProb = FALSE,
weightMap = NULL,
negZero = TRUE,
invWeight = FALSE,
weightDecay = 1) {
XPred <- XPredList[[1]]
XPred$density <- log(XPred$density)
if (length(XPredList) > 1) {
for (j in 2:length(XPredList)) {
closest <- sapply(1:nrow(XPred), function(x)
XPredList[[j]]$density[which.min((XPred$Longitude[x] - XPredList[[j]]$Longitude) ^ 2 +
(XPred$Latitude[x] - XPredList[[j]]$Latitude) ^ 2
)])
closestSD <- sapply(1:nrow(XPred), function(x)
XPredList[[j]]$densitySd[which.min((XPred$Longitude[x] - XPredList[[j]]$Longitude) ^ 2 +
(XPred$Latitude[x] - XPredList[[j]]$Latitude) ^ 2
)])
closestDist <- sapply(1:nrow(XPred), function(x)
min((XPred$Longitude[x] - XPredList[[j]]$Longitude) ^ 2 +
(XPred$Latitude[x] - XPredList[[j]]$Latitude) ^ 2
))
closest[closestDist > sqrt(100)] <- NA
closestSD[closestDist > sqrt(100)] <- NA
XPred$densitySd <- sqrt((XPred$densitySd ^ 2 + exp(XPred$density) ^ 2) * (closestSD ^ 2 + closest ^ 2) -
(exp(XPred$density) ^ 2) * (closest ^ 2))
XPred$density <- XPred$density + log(closest)
}
}
XPred$density <- exp(XPred$density)
if(!is.null(weightMap) & weightProb == TRUE){
closestWeights <- sapply(1:nrow(XPred), function(x)
weightMap$predictions$Est[which.min((XPred$Longitude[x] - weightMap$predictions$Longitude) ^ 2 +
(XPred$Latitude[x] - weightMap$predictions$Latitude) ^ 2
)])
closestWeightsOG <- closestWeights
if(!negZero){
closestWeights[closestWeightsOG<0] <- 0
}
if(invWeight){
closestWeights <- exp(-log(2)/weightDecay * closestWeights)
}
if(negZero){
closestWeights[closestWeightsOG<0] <- 0
}
XPred$density <- XPred$density * closestWeights
XPred$densitySd <- XPred$densitySd * sqrt(closestWeights)
}
XPred$Est <- XPred$density
if(normalize == TRUE){
if(normalType == "1"){
constant <- max(XPred$Est, na.rm = TRUE)
} else {
constant <- sum(XPred$Est, na.rm = TRUE) /
(diff(sort(unique(XPred$Longitude))[1:2]) *
diff(sort(unique(XPred$Latitude))[1:2]))
}
} else {
constant <- 1
}
XPred$Est <- XPred$Est / constant
XPred$Sd <- XPred$densitySd / sqrt(constant)
XPred$IntLower <- pmax(0, XPred$Est - 1.96 * XPred$Sd)
XPred$IntUpper <- XPred$Est + 1.96 * XPred$Sd
return(XPred)
}
createDifferenceMap <- function(XPred1, XPred2, operation = "-") {
if (inherits(XPred2, "numeric") && !inherits(XPred1, "numeric")) {
XPredNew <- XPred1
XPredNew$Est <- XPred2[1]
XPredNew$Sd <- XPred2[2]
XPred2 <- XPredNew
}
if (inherits(XPred1, "numeric") && !inherits(XPred2, "numeric")) {
XPredNew <- XPred2
XPredNew$Est <- XPred1[1]
XPredNew$Sd <- XPred1[2]
XPred1 <- XPredNew
}
if (inherits(XPred1, "numeric") && inherits(XPred2, "numeric")) {
lo <- seq(-180, 180, by = 0.5)
la <- seq(-90, 90, by = 0.5)
coord <- expand.grid(lo, la)
XPred1 <- data.frame(Est = XPred1[1], Sd = XPred1[2],
Longitude = coord[,1], Latitude = coord[,2])
XPred2 <- data.frame(Est = XPred2[1], Sd = XPred2[2],
Longitude = coord[,1], Latitude = coord[,2])
}
closestMean <- sapply(1:nrow(XPred1), function(x)
XPred2$Est[which.min((XPred1$Longitude[x] - XPred2$Longitude) ^ 2 +
(XPred1$Latitude[x] - XPred2$Latitude) ^ 2)])
closestSd <- sapply(1:nrow(XPred1), function(x)
XPred2$Sd[which.min((XPred1$Longitude[x] - XPred2$Longitude) ^ 2 +
(XPred1$Latitude[x] - XPred2$Latitude) ^ 2)])
closestDist <- sapply(1:nrow(XPred1), function(x)
min((XPred1$Longitude[x] - XPred2$Longitude) ^ 2 +
(XPred1$Latitude[x] - XPred2$Latitude) ^ 2
))
closestMean[closestDist > sqrt(2)] <- NA
closestSd[closestDist > sqrt(2)] <- NA
XPred <- XPred1
if(operation == "-"){
XPred$Est <- XPred1$Est - closestMean
XPred$Sd <- sqrt(XPred1$Sd ^ 2 + closestSd ^ 2)
}
if(operation == "+"){
XPred$Est <- XPred1$Est + closestMean
XPred$Sd <- sqrt(XPred1$Sd ^ 2 + closestSd ^ 2)
}
if(operation == "*"){
XPred$Est <- XPred1$Est * closestMean
XPred$Sd <- sqrt(XPred1$Sd ^ 2 * closestSd ^ 2 +
XPred1$Sd ^ 2 * closestMean ^ 2 +
closestSd ^ 2 * XPred1$Est ^ 2)
}
if(operation == "/"){
XPred$Est <- XPred1$Est / closestMean
XPred$Sd <- sqrt(((XPred1$Est ^ 2) / (closestMean ^ 2)) *
((XPred1$Sd ^ 2) / (XPred1$Est ^ 2) + (closestMean ^ 2) / (closestSd ^ 2)))
}
if(operation == "mean"){
XPred$Est <- (XPred1$Est + closestMean) / 2
XPred$Sd <- sqrt(XPred1$Sd ^ 2 + closestSd ^ 2) / sqrt(2)
}
if(operation == "weightedMean"){
VAR <- 1 / (1 / (XPred1$Sd ^ 2) + 1 / (closestSd ^ 2))
XPred$Est <- VAR * (XPred1$Est / (XPred1$Sd ^ 2) + closestMean / (closestSd ^ 2))
XPred$Sd <- sqrt(VAR)
}
if(operation == "weight"){
constant <- mean(closestMean)
closestMean <- closestMean / constant
closestSd <- closestSd / constant
XPred$Est <- XPred1$Est * closestMean
XPred$Sd <- sqrt(XPred1$Sd ^ 2 * closestSd ^ 2 +
XPred1$Sd ^ 2 * closestMean ^ 2 +
closestSd ^ 2 * XPred1$Est ^ 2)
}
if(operation == "pMin"){
XPred$Est <- pmin(XPred1$Est, closestMean, na.rm = T)
XPred1$Est[is.na(XPred1$Est)] <- -Inf
XPred$Sd <- sapply(1:length(XPred1$Sd), function(l) c(XPred1$Sd[l], closestSd[l])[which.min(c(XPred1$Est[l], closestMean[l]))])
XPred$Sd[is.na(XPred$Est)] <- NA
}
if(operation == "pMax"){
XPred$Est <- pmax(XPred1$Est, closestMean, na.rm = T)
XPred1$Est[is.na(XPred1$Est)] <- Inf
XPred$Sd <- sapply(1:length(XPred1$Sd), function(l) c(XPred1$Sd[l], closestSd[l])[which.max(c(XPred1$Est[l], closestMean[l]))])
XPred$Sd[is.na(XPred$Est)] <- NA
}
XPred$IntLower <- XPred$Est - 1.96 * XPred$Sd
XPred$IntUpper <- XPred$Est + 1.96 * XPred$Sd
return(XPred)
}
getMinima <- function(XPredPlot, nMin = 3, minDist = 250, minima = "Min"){
if(minima == "Max"){
XPredPlot$Est <- - 1* XPredPlot$Est
}
mins <- rep(NA, nMin)
mins[1] <- which.min(XPredPlot$Est)
y = 1
dist <- which(sqrt((XPredPlot$Longitude[mins[y]] - XPredPlot$Longitude) ^ 2 +
(XPredPlot$Latitude[mins[y]] - XPredPlot$Latitude) ^ 2) * 111 > minDist)
while(y < nMin & y < 9){
newMin <- which.min(XPredPlot$Est[dist])
newMinVal <- min(XPredPlot$Est[dist], na.rm = TRUE)
if(length(na.omit(newMin)) == 0){
return(na.omit(mins))
}
nearest <-
sqrt((XPredPlot$Longitude[dist][newMin] - XPredPlot$Longitude[dist]) ^ 2 +
(XPredPlot$Latitude[dist][newMin] - XPredPlot$Latitude[dist]) ^ 2)
neighbours <- XPredPlot$Est[dist][which(nearest <= sort(nearest)[5])]
if(newMinVal == min(neighbours, na.rm = TRUE)){
y = y +1
mins[y] = which(newMinVal == XPredPlot$Est)
dist <- intersect(dist, which(sqrt((XPredPlot$Longitude[mins[y]] - XPredPlot$Longitude) ^ 2 +
(XPredPlot$Latitude[mins[y]] - XPredPlot$Latitude) ^ 2) * 111 > minDist))
} else {
dist <- setdiff(dist, which(newMinVal == XPredPlot$Est))
}
if(length(na.omit(dist)) == 0){
return(na.omit(mins))
}
}
return(na.omit(mins))
}
#' Select Cluster Grouping
#'
#' @param data (data.frame) data if \code{cluster == TRUE} it must contain columns "spatial_cluster"
#' and "temporal_group"
#' @param cluster (logical) if TRUE, a column "cluster" is added to data
#' @param clusterResults (integer) either \code{0} or \code{1}, using "temporal_group" or
#' "spatial_cluster" for the column "cluster", respectively.
#'
#' @return (data.frame) data containing the column "cluster" if \code{cluster == TRUE}, else
#' just the input \code{data}
selectClusterGrouping <- function(data, cluster, clusterResults) {
if(cluster & !is.null(data$spatial_cluster)){
if(clusterResults == 0){
data$cluster <- data$temporal_group
} else {
data$cluster <- data$spatial_cluster
}
}
return(data)
}
#' Get Point Color
#'
#' @param data (data.frame) if cluster, different ids in the column "cluster" will receive a different colour
#' @param cluster (logical) if TRUE, return color for cluster
#' @param palName (character) name of the color palette
#' @param pColor single color or vector of colors that is returned if \code{cluster == FALSE}
#'
#' @return single color or vector of colors
getPColor <- function(data, cluster, palName, pColor) {
if (!cluster) return(pColor)
colorRampPalette(brewer.pal(8, palName))(max(data$cluster))[data$cluster]
}
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.