manuscriptAnalysis/old/linearGR_Models_ATS.R

In bletcher/linkedModels: Link movement, growth, and survival models

library(lme4)

# lm to compare against the jags model
# patterns generally match up well

d <- data.frame(
  sampleNumber=ddG[[1]][[2]]$sampleNumber,
  ind=ddG[[1]][[2]]$tag,
  river=ddG[[1]][[2]]$riverOrdered,
  season=ddG[[1]][[2]]$season,
  species=ddG[[1]][[2]]$species,
  isYOY=ddG[[1]][[2]]$isYOYDATA,
  length=ddG[[1]][[2]]$lengthDATAStd,
  grLength=ddG[[1]][[2]]$grLength,
  tempStd=ddG[[1]][[2]]$tempStd,
  flowStd=ddG[[1]][[2]]$flowStd,
  countPStd=ddG[[1]][[2]]$nAllFishPStd,
  nBKT=ddG[[1]][[2]]$nAllFishBySpeciesPStdBKT,
  nBNT=ddG[[1]][[2]]$nAllFishBySpeciesPStdBNT,
  nATS=ddG[[1]][[2]]$nAllFishBySpeciesPStdATS)

d$ATS01 <- ifelse( d$sampleNumber <= 58,1,0 )
d <- d %>% mutate( tempStd2  =tempStd^2, flowStd2 = flowStd^2, countPStd2 = countPStd^2, length2 = length^2)

#speciesIn <- 'bkt'
#d <- d %>% filter(species == speciesIn, !is.na(grLength))

ggplot(d, aes(tempStd, grLength)) +
  geom_point() +
  geom_smooth(method='lm') +
  facet_grid(season ~ river)

ggplot(d, aes(countPStd, grLength)) +
  geom_point() +
  geom_smooth(method='lm') +
  facet_grid(season ~ species)


pairs(d[,c("nBKT","nBNT","nATS","countPStd")])

ggplot(d, aes(nBKT, grLength)) +
  geom_point() +
  geom_smooth(method='lm') +
  facet_grid(season ~ species)
ggplot(d, aes(nBNT, grLength)) +
  geom_point() +
  geom_smooth(method='lm') +
  facet_grid(season ~ species)
ggplot(d, aes(nATS, grLength)) +
  geom_point() +
  geom_smooth(method='lm') +
  facet_grid(season ~ species)


ggplot(d, aes(flowStd, grLength)) +
  geom_point() +
  geom_smooth(method='lm') +
  facet_grid(season ~ species)

ggplot(d, aes(tempStd, grLength, color=countPStd)) +
  geom_point() +
  geom_smooth(method='lm') +
  facet_grid(season ~ river)

# simple lm
mod0 <- lm( grLength ~ as.factor(isYOY) * river * as.factor(season) * tempStd * flowStd * countPStd + tempStd2+flowStd2+countPStd2, data = d )

# lmer"

mod1 <- lmer( grLength ~ as.factor(isYOY)  + as.factor(season) +
                         tempStd + flowStd + countPStd +
                         tempStd2 + flowStd2 + countPStd2 + (1|ind), data = d)

mod2 <- lmer( grLength ~ as.factor(isYOY)  + as.factor(season) +
                length + tempStd + flowStd + countPStd +
                tempStd2 + flowStd2 + countPStd2 + length2 +
                tempStd * flowStd +
                flowStd * countPStd +
                tempStd * countPStd +
                tempStd * flowStd * countPStd +
                (1|ind), data = d)

mod3 <- lmer( grLength ~ as.factor(isYOY)  + as.factor(season) +
                length + tempStd + flowStd + countPStd +
                ATS01 +
                tempStd2 + flowStd2 + countPStd2 + length2 +
                tempStd * flowStd +
                flowStd * countPStd +
                tempStd * countPStd +
                tempStd * flowStd * countPStd +

                (1|ind), data = d)

mod4 <- lmer( grLength ~ as.factor(isYOY)  + as.factor(season) +
                length + tempStd + flowStd + nBKT + nBNT + nATS +
                tempStd2 + flowStd2 + countPStd2 + length2 +
                tempStd * flowStd +
                flowStd * countPStd +
                tempStd * countPStd +
                tempStd * flowStd * countPStd +

                (1|ind), data = d)

AIC(mod1,mod2,mod3,mod4)













# predictions for the data
d$predicted <- predict( mod1 )
ggplot(d, aes(tempStd, predicted, color=countPStd)) +
  geom_point() +
  facet_grid(season ~ river)

ggplot(d, aes(tempStd, predicted, color=flowStd)) +
  geom_point() +
  facet_grid(season ~ river)


# predict over a grid
x <- seq( -2,2,length.out = 5 )

  isYOYData <- rep(1:4, each = nPoints ^ 5)
  seasonData <- rep(1:4, each = nPoints ^ 4)
  riverData <- rep(unique(d$river), each = nPoints ^ 3)
  tempData <- rep(x, each = nPoints ^ 2)
  flowData <- rep(x, each = nPoints ^ 1)
  countData <- rep(x, each = nPoints ^ 0)


predTemplate <- data.frame( isYOY = isYOYData,
                            season = seasonData,
                            river = riverData,
                            countPStd = countData,
                            flowStd =  flowData,
                            tempStd =  tempData
) %>% filter(isYOY %in% 1:2)
predTemplate <- predTemplate %>% mutate( tempStd2  = tempStd^2, flowStd2 = flowStd^2, countPStd2 = countPStd^2)

predTemplate$predicted <- predict( mod1, newdata=predTemplate )

countIn <- 0
isYOYIn <- 2
ggplot(predTemplate %>% filter(countPStd == countIn, isYOY == isYOYIn), aes(tempStd, predicted, color=flowStd)) +
  geom_line(aes(group=flowStd)) +
  geom_point() +
  ylim(-0.2,1) +
  ggtitle(paste(countIn, isYOYIn,speciesIn)) +
  facet_grid( river ~ season )

tempIn <- 0
isYOYIn <- 2
ggplot(predTemplate %>% filter(tempStd == tempIn, isYOY == isYOYIn), aes(countPStd, predicted, color=flowStd)) +
  geom_line(aes(group=flowStd)) +
  geom_point() +
  ylim(-0.2,1) +
  ggtitle(paste(tempIn, isYOYIn,speciesIn)) +
  facet_grid( river ~ season )

tempIn <- 0
flowIn=0
isYOYIn <- 2
ggplot(predTemplate %>% filter(tempStd == tempIn,flowStd==flowIn, isYOY == isYOYIn), aes(countPStd, predicted)) +
  geom_line() +
  geom_point() +
  ylim(-0.2,1) +
  ggtitle(paste(tempIn, isYOYIn,speciesIn)) +
  facet_grid( river ~ season )