inst/extdata/a_scripts/BioGeoBEARS_LaplacesDemon_v1.R
In nmatzke/BioGeoBEARS: BioGeography with Bayesian (and Likelihood) Evolutionary Analysis in R Scripts
require("ape")
require("rexpokit")
require("cladoRcpp")
#######################################################
# LaplacesDemon functions for Bayesian analysis with BioGeoBEARS
#######################################################
#######################################################
# plotlp
#######################################################
#' Plot LaplacesDemon plots after excluding some burnin
#'
#' Takes the objects output by a \code{\link[LaplacesDemon]{LaplacesDemon}} MCMC search, and plots LnP, parameter values, etc.
#'
#' @param Fit The \code{\link[LaplacesDemon]{LaplacesDemon}} output object.
#' @param MyData The \code{\link[LaplacesDemon]{LaplacesDemon}} input data.
#' @param PDF Plot to a PDF? (Recommended!)
#' @param Parms The parameters
#' @param burnfract The burnin fraction
#' @param ... Additional arguments to standard functions
#' @return The plot
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
#'
plotlp <- function(Fit, MyData, PDF=TRUE, Parms=NULL, burnfract=0.0, ...)
{
defaults='
PDF=TRUE
Parms=NULL
burnfract=0.1
'
if (burnfract > 0)
{
# Edit the burnin for the plot
newFit = change_burnin(Fit, MyData, burnfract=0.1, newsamps=NULL)
} else {
newFit = Fit
}
plot_demonoid2(x=newFit, Data=MyData, PDF=PDF, Parms=Parms)
}
# Re-parameterization
# Transform continuous-sampled parameter on -Inf, +Inf to
# a boundedly-sampled parameter
#
#######################################################
# transform_with_logistic
#######################################################
#' Transform continuously-sampled parameter on -Inf, +Inf to a boundedly-sampled parameter
#'
#' \code{\link[LaplacesDemon]{LaplacesDemon}} likes to run its MCMC sampling on a simple number line. Thus, the likelihood
#' function etc. should transform the numbers into the range desired, e.g. 0-1.
#'
#' @param paramval The input value in the range [-Inf, +Inf]
#' @param minval The minimum rescaled value (default 0)
#' @param maxval The maximum rescaled value (default 1)
#' @return \code{paramval}
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite LaplacesDemon_Tutorial
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
#'
#' transform_with_logistic(paramval=-10, minval=0, maxval=5)
#' # NOTE that when minval=0, maxval=1, you have logit!
#'
#' # Transform regular-space parameters to logistic space
#' invlogit(0.5)
#' transform_with_logistic(paramval=0.5, minval=0, maxval=1)
#' transform_with_logistic(paramval=0.5, minval=0, maxval=5)
#'
#' invlogit(0.3)
#' transform_with_logistic(paramval=0.3, minval=0, maxval=1)
#' transform_with_logistic(paramval=0.3, minval=0, maxval=5)
#'
#' # Transform logistic-space parameters back to regular space
#' logit(0.5744425)
#' invtransform_with_logistic(transformed_paramval=0.5744425, minval=0, maxval=1)
#' invtransform_with_logistic(transformed_paramval=3.112297, minval=0, maxval=5)
#'
#' logit(0.5744425)
#' invtransform_with_logistic(transformed_paramval=0.5744425, minval=0, maxval=1)
#' invtransform_with_logistic(transformed_paramval=2.872213, minval=0, maxval=5)
#'
#' # These should transform, then undo the transform
#' invtransform_with_logistic(transformed_paramval=transform_with_logistic(
#' paramval=0.3, minval=0, maxval=1), minval=0, maxval=1)
#' invtransform_with_logistic(transformed_paramval=transform_with_logistic(
#' paramval=0.3, minval=0, maxval=5), minval=0, maxval=5)
#'
transform_with_logistic <- function(paramval, minval=0, maxval=1)
{
transformed_paramval = minval + maxval * (exp(paramval) / (exp(paramval) + 1))
return(transformed_paramval)
}
#
# # Transform regular-space parameters to logistic space
# invlogit(0.5)
# transform_with_logistic(paramval=0.5, minval=0, maxval=1)
# transform_with_logistic(paramval=0.5, minval=0, maxval=5)
#
# invlogit(0.3)
# transform_with_logistic(paramval=0.3, minval=0, maxval=1)
# transform_with_logistic(paramval=0.3, minval=0, maxval=5)
#
# # Transform logistic-space parameters back to regular space
# logit(0.5744425)
# invtransform_with_logistic(transformed_paramval=0.5744425, minval=0, maxval=1)
# invtransform_with_logistic(transformed_paramval=3.112297, minval=0, maxval=5)
#
# logit(0.5744425)
# invtransform_with_logistic(transformed_paramval=0.5744425, minval=0, maxval=1)
# invtransform_with_logistic(transformed_paramval=2.872213, minval=0, maxval=5)
#
# # These should transform, then undo the transform
# invtransform_with_logistic(transformed_paramval=transform_with_logistic(
#' paramval=0.3, minval=0, maxval=1), minval=0, maxval=1)
# invtransform_with_logistic(transformed_paramval=transform_with_logistic(
#' paramval=0.3, minval=0, maxval=5), minval=0, maxval=5)
#
#######################################################
# invtransform_with_logistic
#######################################################
#' Transform rescaled values back to continuously-sampled parameter on -Inf, +Inf to a boundedly-sampled parameter
#'
#' \code{\link[LaplacesDemon]{LaplacesDemon}} likes to run its MCMC sampling on a simple number line. Thus, the likelihood
#' function etc. should transform the numbers into the range desired, e.g. 0-1. invtransform_with_logistic does the reverse transform.
#'
#' @param transformed_paramval The input value in the range [minval, maxval]
#' @param minval The minimum rescaled value (default 0)
#' @param maxval The maximum rescaled value (default 1)
#' @return \code{paramval}
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite LaplacesDemon_Tutorial
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
#'
#' # Transform regular-space parameters to logistic space
#' invlogit(0.5)
#' transform_with_logistic(paramval=0.5, minval=0, maxval=1)
#' transform_with_logistic(paramval=0.5, minval=0, maxval=5)
#'
#' invlogit(0.3)
#' transform_with_logistic(paramval=0.3, minval=0, maxval=1)
#' transform_with_logistic(paramval=0.3, minval=0, maxval=5)
#'
#' # Transform logistic-space parameters back to regular space
#' logit(0.5744425)
#' invtransform_with_logistic(transformed_paramval=0.5744425, minval=0, maxval=1)
#' invtransform_with_logistic(transformed_paramval=3.112297, minval=0, maxval=5)
#'
#' logit(0.5744425)
#' invtransform_with_logistic(transformed_paramval=0.5744425, minval=0, maxval=1)
#' invtransform_with_logistic(transformed_paramval=2.872213, minval=0, maxval=5)
#'
#' # These should transform, then undo the transform
#' invtransform_with_logistic(transformed_paramval=transform_with_logistic(
#' paramval=0.3, minval=0, maxval=1), minval=0, maxval=1)
#' invtransform_with_logistic(transformed_paramval=transform_with_logistic(
#' paramval=0.3, minval=0, maxval=5), minval=0, maxval=5)
#'
invtransform_with_logistic <- function(transformed_paramval, minval=0, maxval=1)
{
# transformed_paramval = minval + maxval * (exp(paramval) / (exp(paramval) + 1))
# (paramval-minval)/maxval = (exp(transformed_paramval) / (exp(transformed_paramval) + 1))
# log((paramval-minval)/maxval) = log(exp(transformed_paramval)) - log((exp(transformed_paramval) + 1))
# log((paramval-minval)/maxval) = transformed_paramval - (transformed_paramval - 1)
# log((paramval-minval)/maxval) = transformed_paramval
# Error check (as in logit())
if ({
any(transformed_paramval < minval)
} || {
any(transformed_paramval > maxval)
})
stop("transformed_paramval must be in [minval, maxval].")
# Value must be between 0 and 1
val01 = (transformed_paramval-minval)/maxval
paramval = log(val01 / (1-val01))
return(paramval)
}
#######################################################
# reparam_LapDem_output
#######################################################
#' Re-parameterize LaplacesDemon MCMC output
#'
#' \code{\link[LaplacesDemon]{LaplacesDemon}} likes to run its MCMC sampling on a simple number line. Thus, the likelihood
#' function etc. should transform the numbers into the range desired, e.g. 0-1.
#'
#' This function transforms the \code{\link[LaplacesDemon]{LaplacesDemon}} output
#'
#' @param Fit The \code{\link[LaplacesDemon]{LaplacesDemon}} output object.
#' @param MyData The \code{\link[LaplacesDemon]{LaplacesDemon}} input data.
#' @param transfun The function to use for the transformation.
#' @return \code{Fit} The transformed MCMC output.
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite LaplacesDemon_Tutorial
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
reparam_LapDem_output <- function(Fit, MyData, transfun=transform_with_logistic)
{
require(LaplacesDemon)
# Safety
origFit = Fit
# Get the data
BioGeoBEARS_model_object = MyData$BioGeoBEARS_model_object
params_table = BioGeoBEARS_model_object@params_table
freeTF = params_table$type == "free"
for (i in 1:ncol(Fit$Posterior1))
{
minval = params_table$min[freeTF][i]
maxval = params_table$max[freeTF][i]
#######################################################
# Posterior1
#######################################################
# Mark the 0.0s
zerosTF = Fit$Posterior1[, i] == 0.0
# Change the rest to regularspace
Fit$Posterior1[, i] = transfun(paramval=Fit$Posterior1[, i], minval, maxval)
# Convert 0.0s back to 0.0
Fit$Posterior1[zerosTF, i] = 0.0
#######################################################
# Posterior2
#######################################################
if ( !is.null(nrow(Fit$Posterior2)) && (nrow(Fit$Posterior2) >= 1))
{
# Mark the 0.0s
zerosTF = Fit$Posterior2[, i] == 0.0
# Change the rest to regularspace
Fit$Posterior2[, i] = transfun(paramval=Fit$Posterior2[, i], minval, maxval)
# Convert 0.0s back to 0.0
Fit$Posterior2[zerosTF, i] = 0.0
} else {
# If it's 1 row...
for (j in 1:length(Fit$Posterior2))
{
zerosTF = Fit$Posterior2[i] == 0.0
# Change the rest to regularspace
Fit$Posterior2[i] = transfun(paramval=Fit$Posterior2[i], minval, maxval)
# Convert 0.0s back to 0.0
if (zerosTF == TRUE)
{
Fit$Posterior2[i] = 0.0
}
}
}
}
#######################################################
# Re-do the summary, from the parameter-space posterior sample
#######################################################
#######################################################
# re-do the summary, copying from LaplacesDemon()
#######################################################
# Posterior1 is the thinned samples, including everything
LIV = length(Fit$Initial.Values)
thinned = Fit$Posterior1
thinned.rows <- nrow(thinned)
Thinning = Fit$Thinning
Dev <- matrix(Fit$Deviance, ncol=1)
Mon <- Fit$Monitor
Num.Mon <- ncol(Mon)
thinned2 = Fit$Posterior2
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
HD <- BMK.Diagnostic(thinned2, batches = 10)
Ind <- 1 * (HD > 0.5)
BurnIn <- thinned.rows
batch.list <- seq(from = 1, to = nrow(thinned2), by = floor(nrow(thinned2)/10))
for (i in 1:9)
{
if (sum(Ind[, i:9]) == 0)
{
BurnIn <- batch.list[i] - 1
break
}
}
Stat.at <- BurnIn + 1
#rm(batch.list, HD, Ind, thinned2)
}
cat("Assessing Thinning and ESS\n")
acf.temp <- matrix(1, trunc(10 * log10(thinned.rows)), LIV)
ESS1 <- Rec.Thin <- rep(1, LIV)
for (j in 1:LIV) {
temp0 <- acf(thinned[, j], lag.max = nrow(acf.temp),
plot = FALSE)
acf.temp[, j] <- abs(temp0$acf[2:{
nrow(acf.temp) + 1
}, , 1])
ESS1[j] <- ESS(thinned[, j])
Rec.Thin[j] <- which(acf.temp[, j] <= 0.1)[1] * Thinning
}
Rec.Thin <- ifelse(is.na(Rec.Thin), nrow(acf.temp), Rec.Thin)
ESS2 <- ESS(Dev)
ESS3 <- ESS(Mon)
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
if (Stat.at < thinned.rows)
{
ESS4 <- ESS(thinned[Stat.at:thinned.rows, ])
ESS5 <- ESS(Dev[Stat.at:thinned.rows, ])
ESS6 <- ESS(Mon[Stat.at:thinned.rows, ])
}
}
cat("Creating Summaries\n")
Num.Mon <- ncol(Mon)
Summ1 <- matrix(NA, LIV, 7, dimnames = list(MyData$parm.names,
c("Mean", "SD", "MCSE", "ESS", "LB", "Median", "UB")))
Summ1[, 1] <- colMeans(thinned)
Summ1[, 2] <- apply(thinned, 2, sd)
Summ1[, 3] <- 0
Summ1[, 4] <- ESS1
Summ1[, 5] <- apply(thinned, 2, quantile, c(0.025), na.rm = TRUE)
Summ1[, 6] <- apply(thinned, 2, quantile, c(0.5), na.rm = TRUE)
Summ1[, 7] <- apply(thinned, 2, quantile, c(0.975), na.rm = TRUE)
for (i in 1:ncol(thinned)) {
temp <- try(MCSE(thinned[, i]), silent = TRUE)
if (!inherits(temp, "try-error"))
Summ1[i, 3] <- temp
else Summ1[i, 3] <- MCSE(thinned[, i], method = "sample.variance")
}
Deviance <- rep(NA, 7)
Deviance[1] <- mean(Dev)
Deviance[2] <- sd(as.vector(Dev))
temp <- try(MCSE(as.vector(Dev)), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(as.vector(Dev), method = "sample.variance")
Deviance[3] <- temp
Deviance[4] <- ESS2
Deviance[5] <- as.numeric(quantile(Dev, probs = 0.025, na.rm = TRUE))
Deviance[6] <- as.numeric(quantile(Dev, probs = 0.5, na.rm = TRUE))
Deviance[7] <- as.numeric(quantile(Dev, probs = 0.975, na.rm = TRUE))
Summ1 <- rbind(Summ1, Deviance)
for (j in 1:Num.Mon) {
Monitor <- rep(NA, 7)
Monitor[1] <- mean(Mon[, j])
Monitor[2] <- sd(as.vector(Mon[, j]))
temp <- try(MCSE(as.vector(Mon[, j])), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(Mon[, j], method = "sample.variance")
Monitor[3] <- temp
Monitor[4] <- ESS3[j]
Monitor[5] <- as.numeric(quantile(Mon[, j], probs = 0.025,
na.rm = TRUE))
Monitor[6] <- as.numeric(quantile(Mon[, j], probs = 0.5,
na.rm = TRUE))
Monitor[7] <- as.numeric(quantile(Mon[, j], probs = 0.975,
na.rm = TRUE))
Summ1 <- rbind(Summ1, Monitor)
rownames(Summ1)[nrow(Summ1)] <- MyData$mon.names[j]
}
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
Summ2 <- matrix(NA, LIV, 7, dimnames = list(MyData$parm.names,
c("Mean", "SD", "MCSE", "ESS", "LB", "Median", "UB")))
if (Stat.at < thinned.rows)
{
thinned2 <- matrix(thinned[Stat.at:thinned.rows, ], thinned.rows -
Stat.at + 1, ncol(thinned))
Dev2 <- matrix(Dev[Stat.at:thinned.rows, ], thinned.rows -
Stat.at + 1, ncol(Dev))
Mon2 <- matrix(Mon[Stat.at:thinned.rows, ], thinned.rows -
Stat.at + 1, ncol(Mon))
Summ2[, 1] <- colMeans(thinned2)
Summ2[, 2] <- apply(thinned2, 2, sd)
Summ2[, 3] <- 0
Summ2[, 4] <- ESS4
Summ2[, 5] <- apply(thinned2, 2, quantile, c(0.025),
na.rm = TRUE)
Summ2[, 6] <- apply(thinned2, 2, quantile, c(0.5), na.rm = TRUE)
Summ2[, 7] <- apply(thinned2, 2, quantile, c(0.975),
na.rm = TRUE)
for (i in 1:ncol(thinned2))
{
temp <- try(MCSE(thinned2[, i]), silent = TRUE)
if (!inherits(temp, "try-error"))
Summ2[i, 3] <- temp
else Summ2[i, 3] <- MCSE(thinned2[, i], method = "sample.variance")
}
Deviance <- rep(NA, 7)
Deviance[1] <- mean(Dev2)
Deviance[2] <- sd(as.vector(Dev2))
temp <- try(MCSE(as.vector(Dev2)), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(as.vector(Dev2), method = "sample.variance")
Deviance[3] <- temp
Deviance[4] <- ESS5
Deviance[5] <- as.numeric(quantile(Dev2, probs = 0.025,
na.rm = TRUE))
Deviance[6] <- as.numeric(quantile(Dev2, probs = 0.5,
na.rm = TRUE))
Deviance[7] <- as.numeric(quantile(Dev2, probs = 0.975,
na.rm = TRUE))
Summ2 <- rbind(Summ2, Deviance)
for (j in 1:Num.Mon)
{
Monitor <- rep(NA, 7)
Monitor[1] <- mean(Mon2[, j])
Monitor[2] <- sd(as.vector(Mon2[, j]))
temp <- try(MCSE(as.vector(Mon[, j])), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(as.vector(Mon[, j]), method = "sample.variance")
Monitor[3] <- temp
Monitor[4] <- ESS6[j]
Monitor[5] <- as.numeric(quantile(Mon2[, j], probs = 0.025,
na.rm = TRUE))
Monitor[6] <- as.numeric(quantile(Mon2[, j], probs = 0.5,
na.rm = TRUE))
Monitor[7] <- as.numeric(quantile(Mon2[, j], probs = 0.975,
na.rm = TRUE))
Summ2 <- rbind(Summ2, Monitor)
rownames(Summ2)[nrow(Summ2)] <- MyData$mon.names[j]
}
}
Fit$Summary2 = Summ2
}
Fit$Summary1 = Summ1
# Update Deviance, DIC1, DIC2 (actually shouldn't change)
Fit$Deviance = as.vector(Dev)
Fit$DIC1 = c(mean(as.vector(Dev)), var(as.vector(Dev))/2,
mean(as.vector(Dev)) + var(as.vector(Dev))/2)
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
if (Stat.at < thinned.rows)
{
Fit$DIC2 = c(mean(as.vector(Dev2)), var(as.vector(Dev2))/2, mean(as.vector(Dev2)) + var(as.vector(Dev2))/2)
} else {
Fit$DIC2 = rep(NA, 3)
}
}
#######################################################
# Find out how much of each parameter equals 0.0
#######################################################
zeros_TF = Fit$Posterior1 == 0.0
Posterior1_iszero = colSums(zeros_TF) / nrow(Fit$Posterior1)
Fit$Posterior1_iszero = Posterior1_iszero
if ( !is.null(nrow(Fit$Posterior2)) && (nrow(Fit$Posterior2) >= 1))
{
zeros_TF = Fit$Posterior2 == 0.0
Posterior2_iszero = colSums(zeros_TF) / nrow(Fit$Posterior2)
Fit$Posterior2_iszero = Posterior2_iszero
}
return(Fit)
}
#######################################################
# reparam_LapDem_output
#######################################################
#' Change the burnin fraction when re-parameterizing LaplacesDemon MCMC output
#'
#' \code{\link[LaplacesDemon]{LaplacesDemon}} likes to run its MCMC sampling on a simple number line. Thus, the likelihood
#' function etc. should transform the numbers into the range desired, e.g. 0-1.
#'
#' This function changes the burnin fraction
#'
#' @param Fit The \code{\link[LaplacesDemon]{LaplacesDemon}} output object.
#' @param MyData The \code{\link[LaplacesDemon]{LaplacesDemon}} input data.
#' @param burnfract The new burnin fraction
#' @param newsamps If you want to subset the samples.
#' @return \code{Fit} The modified MCMC output.
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite LaplacesDemon_Tutorial
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
change_burnin <- function(Fit, MyData, burnfract=0.1, newsamps=NULL)
{
require(LaplacesDemon)
# Safety
origFit = Fit
# Get the data
BioGeoBEARS_model_object = MyData$BioGeoBEARS_model_object
params_table = BioGeoBEARS_model_object@params_table
freeTF = params_table$type == "free"
# Cut the number of rows by burnfract
rowstart = round(burnfract * nrow(Fit$Posterior1)) + 1
rowend = nrow(Fit$Posterior1)
# The NULL is NOT transferred to seq successfully
if (!is.null(newsamps))
{
rownums = seq(from=rowstart, to=rowend, length.out=newsamps)
} else {
rownums = seq(from=rowstart, to=rowend)
}
# Cut the tables down
Fit$Posterior1 = Fit$Posterior1[rownums, ]
Fit$Monitor = Fit$Monitor[rownums, ]
Fit$Deviance = Fit$Deviance[rownums]
# Revise summary also
Fit$Thinned.Samples = nrow(Fit$Posterior1)
Fit$Iterations = nrow(Fit$Posterior1)
# Maybe also do Posterior2 (should be unnecessary)
dothis = FALSE
if (dothis && !is.null(nrow(Fit$Posterior2)) && (nrow(Fit$Posterior2) >= 1))
{
# Cut the number of rows by burnfract
rowstart = round(burnfract * nrow(Fit$Posterior2)) + 1
rowend = nrow(Fit$Posterior2)
# The NULL is transferred to seq successfully
rownums = seq(start=rowstart, end=rowend, length.out=newsamps)
Fit$Posterior2 = Fit$Posterior2[rownums, ]
}
#######################################################
# Re-do the summary, from the parameter-space posterior sample
#######################################################
#######################################################
# re-do the summary, copying from LaplacesDemon()
#######################################################
# Posterior1 is the thinned samples, including everything
LIV = length(Fit$Initial.Values)
thinned = Fit$Posterior1
thinned.rows <- nrow(thinned)
Thinning = Fit$Thinning
Dev <- matrix(Fit$Deviance, ncol=1)
Mon <- Fit$Monitor
Num.Mon <- ncol(Mon)
thinned2 = Fit$Posterior2
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
HD <- BMK.Diagnostic(thinned2, batches = 10)
Ind <- 1 * (HD > 0.5)
BurnIn <- thinned.rows
batch.list <- seq(from = 1, to = nrow(thinned2), by = floor(nrow(thinned2)/10))
for (i in 1:9)
{
if (sum(Ind[, i:9]) == 0)
{
BurnIn <- batch.list[i] - 1
break
}
}
Stat.at <- BurnIn + 1
#rm(batch.list, HD, Ind, thinned2)
}
cat("Assessing Thinning and ESS\n")
acf.temp <- matrix(1, trunc(10 * log10(thinned.rows)), LIV)
ESS1 <- Rec.Thin <- rep(1, LIV)
for (j in 1:LIV) {
temp0 <- acf(thinned[, j], lag.max = nrow(acf.temp),
plot = FALSE)
acf.temp[, j] <- abs(temp0$acf[2:{
nrow(acf.temp) + 1
}, , 1])
ESS1[j] <- ESS(thinned[, j])
Rec.Thin[j] <- which(acf.temp[, j] <= 0.1)[1] * Thinning
}
Rec.Thin <- ifelse(is.na(Rec.Thin), nrow(acf.temp), Rec.Thin)
ESS2 <- ESS(Dev)
ESS3 <- ESS(Mon)
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
if (Stat.at < thinned.rows)
{
ESS4 <- ESS(thinned[Stat.at:thinned.rows, ])
ESS5 <- ESS(Dev[Stat.at:thinned.rows, ])
ESS6 <- ESS(Mon[Stat.at:thinned.rows, ])
}
}
cat("Creating Summaries\n")
Num.Mon <- ncol(Mon)
Summ1 <- matrix(NA, LIV, 7, dimnames = list(MyData$parm.names,
c("Mean", "SD", "MCSE", "ESS", "LB", "Median", "UB")))
Summ1[, 1] <- colMeans(thinned)
Summ1[, 2] <- apply(thinned, 2, sd)
Summ1[, 3] <- 0
Summ1[, 4] <- ESS1
Summ1[, 5] <- apply(thinned, 2, quantile, c(0.025), na.rm = TRUE)
Summ1[, 6] <- apply(thinned, 2, quantile, c(0.5), na.rm = TRUE)
Summ1[, 7] <- apply(thinned, 2, quantile, c(0.975), na.rm = TRUE)
for (i in 1:ncol(thinned)) {
temp <- try(MCSE(thinned[, i]), silent = TRUE)
if (!inherits(temp, "try-error"))
Summ1[i, 3] <- temp
else Summ1[i, 3] <- MCSE(thinned[, i], method = "sample.variance")
}
Deviance <- rep(NA, 7)
Deviance[1] <- mean(Dev)
Deviance[2] <- sd(as.vector(Dev))
temp <- try(MCSE(as.vector(Dev)), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(as.vector(Dev), method = "sample.variance")
Deviance[3] <- temp
Deviance[4] <- ESS2
Deviance[5] <- as.numeric(quantile(Dev, probs = 0.025, na.rm = TRUE))
Deviance[6] <- as.numeric(quantile(Dev, probs = 0.5, na.rm = TRUE))
Deviance[7] <- as.numeric(quantile(Dev, probs = 0.975, na.rm = TRUE))
Summ1 <- rbind(Summ1, Deviance)
for (j in 1:Num.Mon) {
Monitor <- rep(NA, 7)
Monitor[1] <- mean(Mon[, j])
Monitor[2] <- sd(as.vector(Mon[, j]))
temp <- try(MCSE(as.vector(Mon[, j])), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(Mon[, j], method = "sample.variance")
Monitor[3] <- temp
Monitor[4] <- ESS3[j]
Monitor[5] <- as.numeric(quantile(Mon[, j], probs = 0.025,
na.rm = TRUE))
Monitor[6] <- as.numeric(quantile(Mon[, j], probs = 0.5,
na.rm = TRUE))
Monitor[7] <- as.numeric(quantile(Mon[, j], probs = 0.975,
na.rm = TRUE))
Summ1 <- rbind(Summ1, Monitor)
rownames(Summ1)[nrow(Summ1)] <- MyData$mon.names[j]
}
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
Summ2 <- matrix(NA, LIV, 7, dimnames = list(MyData$parm.names,
c("Mean", "SD", "MCSE", "ESS", "LB", "Median", "UB")))
if (Stat.at < thinned.rows)
{
thinned2 <- matrix(thinned[Stat.at:thinned.rows, ], thinned.rows -
Stat.at + 1, ncol(thinned))
Dev2 <- matrix(Dev[Stat.at:thinned.rows, ], thinned.rows -
Stat.at + 1, ncol(Dev))
Mon2 <- matrix(Mon[Stat.at:thinned.rows, ], thinned.rows -
Stat.at + 1, ncol(Mon))
Summ2[, 1] <- colMeans(thinned2)
Summ2[, 2] <- apply(thinned2, 2, sd)
Summ2[, 3] <- 0
Summ2[, 4] <- ESS4
Summ2[, 5] <- apply(thinned2, 2, quantile, c(0.025),
na.rm = TRUE)
Summ2[, 6] <- apply(thinned2, 2, quantile, c(0.5), na.rm = TRUE)
Summ2[, 7] <- apply(thinned2, 2, quantile, c(0.975),
na.rm = TRUE)
for (i in 1:ncol(thinned2))
{
temp <- try(MCSE(thinned2[, i]), silent = TRUE)
if (!inherits(temp, "try-error"))
Summ2[i, 3] <- temp
else Summ2[i, 3] <- MCSE(thinned2[, i], method = "sample.variance")
}
Deviance <- rep(NA, 7)
Deviance[1] <- mean(Dev2)
Deviance[2] <- sd(as.vector(Dev2))
temp <- try(MCSE(as.vector(Dev2)), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(as.vector(Dev2), method = "sample.variance")
Deviance[3] <- temp
Deviance[4] <- ESS5
Deviance[5] <- as.numeric(quantile(Dev2, probs = 0.025,
na.rm = TRUE))
Deviance[6] <- as.numeric(quantile(Dev2, probs = 0.5,
na.rm = TRUE))
Deviance[7] <- as.numeric(quantile(Dev2, probs = 0.975,
na.rm = TRUE))
Summ2 <- rbind(Summ2, Deviance)
for (j in 1:Num.Mon)
{
Monitor <- rep(NA, 7)
Monitor[1] <- mean(Mon2[, j])
Monitor[2] <- sd(as.vector(Mon2[, j]))
temp <- try(MCSE(as.vector(Mon[, j])), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(as.vector(Mon[, j]), method = "sample.variance")
Monitor[3] <- temp
Monitor[4] <- ESS6[j]
Monitor[5] <- as.numeric(quantile(Mon2[, j], probs = 0.025,
na.rm = TRUE))
Monitor[6] <- as.numeric(quantile(Mon2[, j], probs = 0.5,
na.rm = TRUE))
Monitor[7] <- as.numeric(quantile(Mon2[, j], probs = 0.975,
na.rm = TRUE))
Summ2 <- rbind(Summ2, Monitor)
rownames(Summ2)[nrow(Summ2)] <- MyData$mon.names[j]
}
}
Fit$Summary2 = Summ2
}
Fit$Summary1 = Summ1
# Update Deviance, DIC1, DIC2 (actually shouldn't change)
Fit$Deviance = as.vector(Dev)
Fit$DIC1 = c(mean(as.vector(Dev)), var(as.vector(Dev))/2,
mean(as.vector(Dev)) + var(as.vector(Dev))/2)
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
if (Stat.at < thinned.rows)
{
Fit$DIC2 = c(mean(as.vector(Dev2)), var(as.vector(Dev2))/2, mean(as.vector(Dev2)) + var(as.vector(Dev2))/2)
} else {
Fit$DIC2 = rep(NA, 3)
}
}
#######################################################
# Find out how much of each parameter equals 0.0
#######################################################
zeros_TF = Fit$Posterior1 == 0.0
Posterior1_iszero = colSums(zeros_TF) / nrow(Fit$Posterior1)
Fit$Posterior1_iszero = Posterior1_iszero
if ( !is.null(nrow(Fit$Posterior2)) && (nrow(Fit$Posterior2) >= 1))
{
zeros_TF = Fit$Posterior2 == 0.0
Posterior2_iszero = colSums(zeros_TF) / nrow(Fit$Posterior2)
Fit$Posterior2_iszero = Posterior2_iszero
}
return(Fit)
}
#######################################################
# Bayesian models, for LaplacesDemon
#######################################################
# params = c(0.1, 0.1, 0.3)
# Calculate the log posterior probability
#######################################################
# LapDem_Model_calc_logPP
#######################################################
#' The function calculating posterior probability of parameter values
#'
#' What it says
#'
#' @param params the parameter values
#' @param MyData The \code{\link[LaplacesDemon]{LaplacesDemon}} input data.
#' @return \code{Modelout} The MCMC output.
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite LaplacesDemon_Tutorial
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
LapDem_Model_calc_logPP <- function(params, MyData)
{
require(LaplacesDemon) # for logit, invlogit
orig_params = params
# Get the data
tr = MyData$tr
tipranges = MyData$tipranges
BioGeoBEARS_model_object = MyData$BioGeoBEARS_model_object
params_table = BioGeoBEARS_model_object@params_table
freeTF = params_table$type == "free"
# NO: Convert the params (range -Inf -> +Inf) to the min/max values
# convert anything outside the range, to min/max, using the interval function
params
for (p in 1:length(params))
{
# Check for parameters being set to 0 by RJMCMC
if (params[p] == 0.0)
{
params[p] = 0.0
} else {
# Inverse logit of param
# params[p] = invlogit(params[p])
# Reset things outside the boundary to (min-max)
minval = params_table$min[freeTF][p]
maxval = params_table$max[freeTF][p]
params[p] = transform_with_logistic(paramval=params[p], minval, maxval)
}
}
params
# NO, INTERVAL() CRASHES with mapply or anything else
# d = 0.1
# e = 0.1
# j = 0.3
# params = c(d,e,j)
# #params = c(d,e)
#
# params_table = BioGeoBEARS_model_object@params_table
# freeTF = params_table$type == "free"
#
#
# # Re-parameterize/constrain with LaplacesDemon:::interval()
# params_not_zero_TF = params != 0.0
# minvals = params_table$min[freeTF][params_not_zero_TF]
# maxvals = params_table$min[freeTF][params_not_zero_TF]
# mapply(
# Input the current parameters into the BioGeoBEARS model object
BioGeoBEARS_model_object = params_into_BioGeoBEARS_model_object(BioGeoBEARS_model_object, params)
# Update the dependent parameters accordingly
BioGeoBEARS_model_object = calc_linked_params_BioGeoBEARS_model_object(BioGeoBEARS_model_object)
# Input the basics for calc_loglike_for_optim()
tip_condlikes_of_data_on_each_state = MyData$tip_condlikes_of_data_on_each_state
print_optim = MyData$print_optim
areas_list = MyData$areas_list
states_list = MyData$states_list
force_sparse = MyData$force_sparse
cluster_already_open = MyData$cluster_already_open
### Parameters
params
### Log of Prior Densities (these are all uniform)
#beta.prior <- sum(dnormv(beta, 0, 1000, log=TRUE))
#sigma.prior <- dhalfcauchy(sigma, 25, log=TRUE)
params_table = BioGeoBEARS_model_object@params_table
freeTF = params_table$type == "free"
rownums = (1:nrow(params_table))[freeTF]
LnPrior = 0
for (rn in rownums)
{
# Assume a Unif(min, max) prior for all variables
LnPrior = LnPrior + dunif(x=params_table$est[rn], min=params_table$min[rn], max=params_table$max[rn], log=TRUE)
# Assume a very flat, but normal, prior, for all variables
# (This will avoid -- NAH, doesn't work cause variables are in regular flat space, not logit space
# NO DOESNT WORK
#LnPrior = LnPrior + rnorm(x=params_table$est[rn], min=params_table$min[rn], max=params_table$max[rn], log=TRUE)
# Try a beta prior with shape1 & shape2 close to 1.00001
plotthis = '
x = seq(0,1,0.01)
y=dbeta(x, shape1=1.01, shape2=1.01)
plot(x,y)
dbeta(x, shape1=1.01, shape2=1.01, log=TRUE)
'
# if ( (rownames(params_table)[rn] == "d") || (rownames(params_table)[rn] == "e"))
# {
# # beta prior for d and e
# LnPrior = LnPrior + dbeta(x=params_table$est[rn], shape1=1.000001, shape2=1.000001, log=TRUE)
# } else {
# # unif prior for j etc.
# LnPrior = LnPrior + dunif(x=params_table$est[rn], min=params_table$min[rn], max=params_table$max[rn], log=TRUE)
# }
#
}
### Log-Likelihood
# mus are the coefficients of the betas
# LL is the sum of the logs of the deviance/densities under the model
#mu <- tcrossprod(Data$X, t(beta))
#LL <- sum(dnorm(Data$y, mu, sigma, log=TRUE))
LnL = calc_loglike_for_optim(params, BioGeoBEARS_model_object, phy=tr, tip_condlikes_of_data_on_each_state, print_optim, areas_list=areas_list, states_list=states_list, force_sparse=force_sparse, cluster_already_open=cluster_already_open)
### Log-Posterior
LnP <- LnL + LnPrior
### yhat = data simulated under these model parameters, for posterior predictive sampling/testing
### (I think)
# yhat = rnorm(length(mu), mu, sigma)
yhat = NA
# Convert the params back to (range -Inf -> +Inf) to the min/max values
# for (p in 1:length(params))
# {
# # Scale to logit
# minval = params_table$min[freeTF][p]
# maxval = params_table$max[freeTF][p]
#
#
# if (params[p] == 0.0)
# {
# # Convert to -70 in logit space, to keep parameter "off the bottom" of precision
# params[p] = -70.0
# } else {
# # Don't subtract minval, as that results in 0, which results in -Inf under logit
# params[p] = (params[p] - minval) / (maxval-minval)
#
# # logit of param
# params[p] = logit(params[p])
# }
# }
### Model output to return
Modelout <- list(LP=LnP, Dev=-2*LnL, Monitor=c(LnP,LnL,LnPrior), yhat=yhat, parm=orig_params)
return(Modelout)
}
# Calculate the log posterior probability
# This version attempts rescaling rather than re-writing parameters...
#######################################################
# LapDem_Model_calc_logPP_orig
#######################################################
#' The function calculating posterior probability of parameter values
#'
#' What it says
#'
#' @param params the parameter values
#' @param MyData The \code{\link[LaplacesDemon]{LaplacesDemon}} input data.
#' @return \code{Modelout} The MCMC output.
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite LaplacesDemon_Tutorial
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
LapDem_Model_calc_logPP_orig <- function(params, MyData)
{
require(LaplacesDemon) # for logit, invlogit
# Get the data
tr = MyData$tr
tipranges = MyData$tipranges
BioGeoBEARS_model_object = MyData$BioGeoBEARS_model_object
params_table = BioGeoBEARS_model_object@params_table
freeTF = params_table$type == "free"
# Convert the params (range -Inf -> +Inf) to the min/max values
for (p in 1:length(params))
{
# Check for parameters being set to 0 by RJMCMC
if (params[p] == 0.0)
{
params[p] = 0.0
} else {
# Inverse logit of param
params[p] = invlogit(params[p])
# Scale to (min-max)
minval = params_table$min[freeTF][p]
maxval = params_table$max[freeTF][p]
params[p] = minval + (maxval-minval) * params[p]
}
}
# Input the current parameters into the BioGeoBEARS model object
BioGeoBEARS_model_object = params_into_BioGeoBEARS_model_object(BioGeoBEARS_model_object, params)
# Update the dependent parameters accordingly
BioGeoBEARS_model_object = calc_linked_params_BioGeoBEARS_model_object(BioGeoBEARS_model_object)
# Input the basics for calc_loglike_for_optim()
tip_condlikes_of_data_on_each_state = MyData$tip_condlikes_of_data_on_each_state
print_optim = MyData$print_optim
areas_list = MyData$areas_list
states_list = MyData$states_list
force_sparse = MyData$force_sparse
cluster_already_open = MyData$cluster_already_open
### Parameters
params
### Log of Prior Densities (these are all uniform)
#beta.prior <- sum(dnormv(beta, 0, 1000, log=TRUE))
#sigma.prior <- dhalfcauchy(sigma, 25, log=TRUE)
params_table = BioGeoBEARS_model_object@params_table
freeTF = params_table$type == "free"
rownums = (1:nrow(params_table))[freeTF]
LnPrior = 0
for (rn in rownums)
{
# Assume a Unif(min, max) prior for all variables
#LnPrior = LnPrior + dunif(x=params_table$est[rn], min=params_table$min[rn], max=params_table$max[rn], log=TRUE)
# Assume a very flat, but normal, prior, for all variables
# (This will avoid -- NAH, doesn't work cause variables are in regular flat space, not logit space
# NO DOESNT WORK
#LnPrior = LnPrior + rnorm(x=params_table$est[rn], min=params_table$min[rn], max=params_table$max[rn], log=TRUE)
# Try a beta prior with shape1 & shape2 close to 1.00001
plotthis = '
x = seq(0,1,0.01)
y=dbeta(x, shape1=1.01, shape2=1.01)
plot(x,y)
dbeta(x, shape1=1.01, shape2=1.01, log=TRUE)
'
if ( (rownames(params_table)[rn] == "d") || (rownames(params_table)[rn] == "e"))
{
# beta prior for d and e
LnPrior = LnPrior + dbeta(x=params_table$est[rn], shape1=1.000001, shape2=1.000001, log=TRUE)
} else {
# unif prior for j etc.
LnPrior = LnPrior + dunif(x=params_table$est[rn], min=params_table$min[rn], max=params_table$max[rn], log=TRUE)
}
}
### Log-Likelihood
# mus are the coefficients of the betas
# LL is the sum of the logs of the deviance/densities under the model
#mu <- tcrossprod(Data$X, t(beta))
#LL <- sum(dnorm(Data$y, mu, sigma, log=TRUE))
LnL = calc_loglike_for_optim(params, BioGeoBEARS_model_object, phy=tr, tip_condlikes_of_data_on_each_state, print_optim, areas_list=areas_list, states_list=states_list, force_sparse=force_sparse, cluster_already_open=cluster_already_open)
### Log-Posterior
LnP <- LnL + LnPrior
### yhat = data simulated under these model parameters, for posterior predictive sampling/testing
### (I think)
# yhat = rnorm(length(mu), mu, sigma)
yhat = NA
# Convert the params back to (range -Inf -> +Inf) to the min/max values
for (p in 1:length(params))
{
# Scale to logit
minval = params_table$min[freeTF][p]
maxval = params_table$max[freeTF][p]
if (params[p] == 0.0)
{
# Convert to -70 in logit space, to keep parameter "off the bottom" of precision
params[p] = -70.0
} else {
# Don't subtract minval, as that results in 0, which results in -Inf under logit
params[p] = (params[p] - minval) / (maxval-minval)
# logit of param
params[p] = logit(params[p])
}
}
### Model output to return
Modelout <- list(LP=LnP, Dev=-2*LnL, Monitor=c(LnP,LnL,LnPrior), yhat=yhat, parm=params)
return(Modelout)
}
#######################################################
# PLotting functions
#######################################################
#LaplacesDemon:::plot.demonoid <- function (x, BurnIn = 0, Data = NULL, PDF = FALSE, Parms = NULL,
# ...)
#######################################################
# plot_demonoid2
#######################################################
#' Plot LaplacesDemon MCMC output
#'
#' Modified version of \code{\link[LaplacesDemon]{plot.demonoid}} to plot to PDF.
#'
#' @param x The MCMC output
#' @param BurnIn The burnin amount
#' @param Data The \code{\link[LaplacesDemon]{LaplacesDemon}} input data.
#' @param PDF If TRUE, plot to PDF.
#' @param Parms The parameters.
#' @param ... Additional arguments to standard functions
#' @return Nothing.
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite LaplacesDemon_Tutorial
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
plot_demonoid2 <- function (x, BurnIn = 0, Data = NULL, PDF = FALSE, Parms = NULL,
...)
{
greyval1 = 0.75
greyval1 = rgb(greyval1, greyval1, greyval1)
greyval2 = 0.35
greyval2 = rgb(greyval2, greyval2, greyval2)
if (missing(x))
stop("The x argument is required.")
if (class(x) != "demonoid")
stop("x must be of class demonoid.")
if (is.null(Data))
stop("The Data argument is NULL.")
if (BurnIn >= nrow(x$Posterior1))
BurnIn <- 0
Stat.at <- BurnIn + 1
if (is.null(Parms)) {
Posterior <- x$Posterior1
}
else {
Parms <- sub("\\[", "\\\\[", Parms)
Parms <- sub("\\]", "\\\\]", Parms)
Parms <- sub("\\.", "\\\\.", Parms)
if (length(grep(Parms[1], colnames(x$Posterior1))) ==
0)
stop("Parameter in Parms does not exist.")
keepcols <- grep(Parms[1], colnames(x$Posterior1))
if (length(Parms) > 1) {
for (i in 2:length(Parms)) {
if (length(grep(Parms[i], colnames(x$Posterior1))) ==
0)
stop("Parameter in Parms does not exist.")
keepcols <- c(keepcols, grep(Parms[i], colnames(x$Posterior1)))
}
}
Posterior <- as.matrix(x$Posterior1[, keepcols])
colnames(Posterior) <- colnames(x$Posterior1)[keepcols]
}
if (PDF == TRUE) {
pdf("LaplacesDemon.Plots.pdf")
par(mfrow = c(3, 3))
}
else {
par(mfrow = c(3, 3), ask = TRUE)
}
for (j in 1:ncol(Posterior)) {
plot(Stat.at:x$Thinned.Samples, Posterior[Stat.at:x$Thinned.Samples,
j], type = "l", xlab = "Generations", ylab = "Value",
main = colnames(Posterior)[j])
panel.smooth(Stat.at:x$Thinned.Samples, Posterior[Stat.at:x$Thinned.Samples,
j], pch = "")
###########################################
# Plot the parameter values
###########################################
numbreaks = 30
vals = pretty(Posterior[Stat.at:x$Thinned.Samples, j])
breakpoints = seq(from=min(vals), to=max(vals), length.out=numbreaks)
# Get values for density plot
histres2 = hist(x=Posterior[Stat.at:x$Thinned.Samples, j], breaks=100, plot=FALSE)
histres = hist(x=Posterior[Stat.at:x$Thinned.Samples, j], breaks=breakpoints, freq=FALSE, add=FALSE, xlab="Value", ylab="Density",
col=greyval1, border=greyval2, main=colnames(Posterior)[j])
lines(density(unlist(mapply(rep,histres2$mids,histres2$counts)), adjust=1), col="black", lwd=1.5)
#polygon(density(Posterior[Stat.at:x$Thinned.Samples, j]), col = "white", border = "black")
abline(v = 0, col = "red", lty = 2)
if (!is.constant(Posterior[Stat.at:x$Thinned.Samples,
j])) {
z <- acf(Posterior[Stat.at:x$Thinned.Samples, j],
plot = FALSE)
se <- 1/sqrt(length(Posterior[Stat.at:x$Thinned.Samples,
j]))
plot(z$lag, z$acf, ylim = c(min(z$acf, -2 * se),
1), type = "h", main = colnames(Posterior)[j],
xlab = "Lag", ylab = "Correlation")
abline(h = (2 * se), col = "red", lty = 2)
abline(h = (-2 * se), col = "red", lty = 2)
}
else {
plot(0, 0, main = paste(colnames(Posterior)[j], "is a constant."))
}
}
plot(Stat.at:length(x$Deviance), x$Deviance[Stat.at:length(x$Deviance)],
type = "l", xlab = "Iterations", ylab = "Value", main = "Deviance")
panel.smooth(Stat.at:length(x$Deviance), x$Deviance[Stat.at:length(x$Deviance)],
pch = "")
#plot(density(x$Deviance[Stat.at:length(x$Deviance)]), xlab = "Value",
# main = "Deviance")
#polygon(density(x$Deviance[Stat.at:length(x$Deviance)]),
# col = "black", border = "black")
# Get values for density plot
###########################################
# Plot the Deviance values
###########################################
numbreaks = 30
vals = pretty(x$Deviance[Stat.at:length(x$Deviance)])
breakpoints = seq(from=min(vals), to=max(vals), length.out=numbreaks)
histres2 = hist(x=x$Deviance[Stat.at:length(x$Deviance)], breaks=100, plot=FALSE)
histres = hist(x=x$Deviance[Stat.at:length(x$Deviance)], breaks=breakpoints, freq=FALSE, add=FALSE, xlab="Value", ylab="Density",
col=greyval1, border=greyval2, main="Deviance")
lines(density(unlist(mapply(rep,histres2$mids,histres2$counts)), adjust=1), col="black", lwd=1.5)
abline(v = 0, col = "red", lty = 2)
if (!is.constant(x$Deviance[Stat.at:length(x$Deviance)])) {
z <- acf(x$Deviance[Stat.at:length(x$Deviance)], plot = FALSE)
se <- 1/sqrt(length(x$Deviance[Stat.at:length(x$Deviance)]))
plot(z$lag, z$acf, ylim = c(min(z$acf, -2 * se), 1),
type = "h", main = "Deviance", xlab = "Lag", ylab = "Correlation")
abline(h = (2 * se), col = "red", lty = 2)
abline(h = (-2 * se), col = "red", lty = 2)
}
else {
plot(0, 0, main = "Deviance is a constant.")
}
if (is.vector(x$Monitor)) {
J <- 1
nn <- length(x$Monitor)
}
else if (is.matrix(x$Monitor)) {
J <- ncol(x$Monitor)
nn <- nrow(x$Monitor)
}
for (j in 1:J) {
plot(Stat.at:nn, x$Monitor[Stat.at:nn, j], type = "l",
xlab = "Iterations", ylab = "Value", main = Data$mon.names[j])
panel.smooth(Stat.at:nn, x$Monitor[Stat.at:nn, j], pch = "")
#plot(density(x$Monitor[Stat.at:nn, j]), xlab = "Value",
# main = Data$mon.names[j])
#polygon(density(x$Monitor[Stat.at:nn, j]), col = "black",
# border = "black")
###########################################
# Plot the LP, LnL, LnPrior values
###########################################
numbreaks = 30
vals = pretty(x$Monitor[Stat.at:nn, j])
breakpoints = seq(from=min(vals), to=max(vals), length.out=numbreaks)
histres2 = hist(x=x$Monitor[Stat.at:nn, j], breaks=100, plot=FALSE)
histres = hist(x=x$Monitor[Stat.at:nn, j], breaks=breakpoints, freq=FALSE, add=FALSE, xlab="Value", ylab="Density",
col=greyval1, border=greyval2, main=Data$mon.names[j])
lines(density(unlist(mapply(rep,histres2$mids,histres2$counts)), adjust=1), col="black", lwd=1.5)
abline(v = 0, col = "red", lty = 2)
if (!is.constant(x$Monitor[Stat.at:nn, j])) {
z <- acf(x$Monitor[Stat.at:nn, j], plot = FALSE)
se <- 1/sqrt(length(x$Monitor[Stat.at:nn, j]))
plot(z$lag, z$acf, ylim = c(min(z$acf, -2 * se),
1), type = "h", main = Data$mon.names[j], xlab = "Lag",
ylab = "Correlation")
abline(h = (2 * se), col = "red", lty = 2)
abline(h = (-2 * se), col = "red", lty = 2)
}
else {
plot(0, 0, main = paste(Data$mon.names[j], "is a constant."))
}
}
if (nrow(x$CovarDHis) > 1) {
if (x$Algorithm %in% c("Adaptive Hamiltonian Monte Carlo",
"Hamiltonian Monte Carlo with Dual-Averaging", "No-U-Turn Sampler")) {
plot(x$CovarDHis[, 1], type = "l", xlab = "Adaptations",
ylab = expression(epsilon))
}
else {
Diff <- abs(diff(x$CovarDHis))
adaptchange <- matrix(NA, nrow(Diff), 3)
for (i in 1:nrow(Diff)) {
adaptchange[i, 1:3] <- as.vector(quantile(Diff[i,
], probs = c(0.025, 0.5, 0.975)))
}
plot(adaptchange[, 2], ylim = c(min(adaptchange),
max(adaptchange)), type = "l", col = "red", xlab = "Adaptations",
ylab = "Absolute Difference", main = "Proposal Variance",
sub = "Median=Red, 95% Bounds=Gray")
lines(adaptchange[, 1], col = "gray")
lines(adaptchange[, 3], col = "gray")
}
}
if (PDF == TRUE)
dev.off()
}
nmatzke/BioGeoBEARS documentation built on April 11, 2024, 2:16 a.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.
require("ape")
require("rexpokit")
require("cladoRcpp")
#######################################################
# LaplacesDemon functions for Bayesian analysis with BioGeoBEARS
#######################################################
#######################################################
# plotlp
#######################################################
#' Plot LaplacesDemon plots after excluding some burnin
#'
#' Takes the objects output by a \code{\link[LaplacesDemon]{LaplacesDemon}} MCMC search, and plots LnP, parameter values, etc.
#'
#' @param Fit The \code{\link[LaplacesDemon]{LaplacesDemon}} output object.
#' @param MyData The \code{\link[LaplacesDemon]{LaplacesDemon}} input data.
#' @param PDF Plot to a PDF? (Recommended!)
#' @param Parms The parameters
#' @param burnfract The burnin fraction
#' @param ... Additional arguments to standard functions
#' @return The plot
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
#'
plotlp <- function(Fit, MyData, PDF=TRUE, Parms=NULL, burnfract=0.0, ...)
{
defaults='
PDF=TRUE
Parms=NULL
burnfract=0.1
'
if (burnfract > 0)
{
# Edit the burnin for the plot
newFit = change_burnin(Fit, MyData, burnfract=0.1, newsamps=NULL)
} else {
newFit = Fit
}
plot_demonoid2(x=newFit, Data=MyData, PDF=PDF, Parms=Parms)
}
# Re-parameterization
# Transform continuous-sampled parameter on -Inf, +Inf to
# a boundedly-sampled parameter
#
#######################################################
# transform_with_logistic
#######################################################
#' Transform continuously-sampled parameter on -Inf, +Inf to a boundedly-sampled parameter
#'
#' \code{\link[LaplacesDemon]{LaplacesDemon}} likes to run its MCMC sampling on a simple number line. Thus, the likelihood
#' function etc. should transform the numbers into the range desired, e.g. 0-1.
#'
#' @param paramval The input value in the range [-Inf, +Inf]
#' @param minval The minimum rescaled value (default 0)
#' @param maxval The maximum rescaled value (default 1)
#' @return \code{paramval}
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite LaplacesDemon_Tutorial
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
#'
#' transform_with_logistic(paramval=-10, minval=0, maxval=5)
#' # NOTE that when minval=0, maxval=1, you have logit!
#'
#' # Transform regular-space parameters to logistic space
#' invlogit(0.5)
#' transform_with_logistic(paramval=0.5, minval=0, maxval=1)
#' transform_with_logistic(paramval=0.5, minval=0, maxval=5)
#'
#' invlogit(0.3)
#' transform_with_logistic(paramval=0.3, minval=0, maxval=1)
#' transform_with_logistic(paramval=0.3, minval=0, maxval=5)
#'
#' # Transform logistic-space parameters back to regular space
#' logit(0.5744425)
#' invtransform_with_logistic(transformed_paramval=0.5744425, minval=0, maxval=1)
#' invtransform_with_logistic(transformed_paramval=3.112297, minval=0, maxval=5)
#'
#' logit(0.5744425)
#' invtransform_with_logistic(transformed_paramval=0.5744425, minval=0, maxval=1)
#' invtransform_with_logistic(transformed_paramval=2.872213, minval=0, maxval=5)
#'
#' # These should transform, then undo the transform
#' invtransform_with_logistic(transformed_paramval=transform_with_logistic(
#' paramval=0.3, minval=0, maxval=1), minval=0, maxval=1)
#' invtransform_with_logistic(transformed_paramval=transform_with_logistic(
#' paramval=0.3, minval=0, maxval=5), minval=0, maxval=5)
#'
transform_with_logistic <- function(paramval, minval=0, maxval=1)
{
transformed_paramval = minval + maxval * (exp(paramval) / (exp(paramval) + 1))
return(transformed_paramval)
}
#
# # Transform regular-space parameters to logistic space
# invlogit(0.5)
# transform_with_logistic(paramval=0.5, minval=0, maxval=1)
# transform_with_logistic(paramval=0.5, minval=0, maxval=5)
#
# invlogit(0.3)
# transform_with_logistic(paramval=0.3, minval=0, maxval=1)
# transform_with_logistic(paramval=0.3, minval=0, maxval=5)
#
# # Transform logistic-space parameters back to regular space
# logit(0.5744425)
# invtransform_with_logistic(transformed_paramval=0.5744425, minval=0, maxval=1)
# invtransform_with_logistic(transformed_paramval=3.112297, minval=0, maxval=5)
#
# logit(0.5744425)
# invtransform_with_logistic(transformed_paramval=0.5744425, minval=0, maxval=1)
# invtransform_with_logistic(transformed_paramval=2.872213, minval=0, maxval=5)
#
# # These should transform, then undo the transform
# invtransform_with_logistic(transformed_paramval=transform_with_logistic(
#' paramval=0.3, minval=0, maxval=1), minval=0, maxval=1)
# invtransform_with_logistic(transformed_paramval=transform_with_logistic(
#' paramval=0.3, minval=0, maxval=5), minval=0, maxval=5)
#
#######################################################
# invtransform_with_logistic
#######################################################
#' Transform rescaled values back to continuously-sampled parameter on -Inf, +Inf to a boundedly-sampled parameter
#'
#' \code{\link[LaplacesDemon]{LaplacesDemon}} likes to run its MCMC sampling on a simple number line. Thus, the likelihood
#' function etc. should transform the numbers into the range desired, e.g. 0-1. invtransform_with_logistic does the reverse transform.
#'
#' @param transformed_paramval The input value in the range [minval, maxval]
#' @param minval The minimum rescaled value (default 0)
#' @param maxval The maximum rescaled value (default 1)
#' @return \code{paramval}
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite LaplacesDemon_Tutorial
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
#'
#' # Transform regular-space parameters to logistic space
#' invlogit(0.5)
#' transform_with_logistic(paramval=0.5, minval=0, maxval=1)
#' transform_with_logistic(paramval=0.5, minval=0, maxval=5)
#'
#' invlogit(0.3)
#' transform_with_logistic(paramval=0.3, minval=0, maxval=1)
#' transform_with_logistic(paramval=0.3, minval=0, maxval=5)
#'
#' # Transform logistic-space parameters back to regular space
#' logit(0.5744425)
#' invtransform_with_logistic(transformed_paramval=0.5744425, minval=0, maxval=1)
#' invtransform_with_logistic(transformed_paramval=3.112297, minval=0, maxval=5)
#'
#' logit(0.5744425)
#' invtransform_with_logistic(transformed_paramval=0.5744425, minval=0, maxval=1)
#' invtransform_with_logistic(transformed_paramval=2.872213, minval=0, maxval=5)
#'
#' # These should transform, then undo the transform
#' invtransform_with_logistic(transformed_paramval=transform_with_logistic(
#' paramval=0.3, minval=0, maxval=1), minval=0, maxval=1)
#' invtransform_with_logistic(transformed_paramval=transform_with_logistic(
#' paramval=0.3, minval=0, maxval=5), minval=0, maxval=5)
#'
invtransform_with_logistic <- function(transformed_paramval, minval=0, maxval=1)
{
# transformed_paramval = minval + maxval * (exp(paramval) / (exp(paramval) + 1))
# (paramval-minval)/maxval = (exp(transformed_paramval) / (exp(transformed_paramval) + 1))
# log((paramval-minval)/maxval) = log(exp(transformed_paramval)) - log((exp(transformed_paramval) + 1))
# log((paramval-minval)/maxval) = transformed_paramval - (transformed_paramval - 1)
# log((paramval-minval)/maxval) = transformed_paramval
# Error check (as in logit())
if ({
any(transformed_paramval < minval)
} || {
any(transformed_paramval > maxval)
})
stop("transformed_paramval must be in [minval, maxval].")
# Value must be between 0 and 1
val01 = (transformed_paramval-minval)/maxval
paramval = log(val01 / (1-val01))
return(paramval)
}
#######################################################
# reparam_LapDem_output
#######################################################
#' Re-parameterize LaplacesDemon MCMC output
#'
#' \code{\link[LaplacesDemon]{LaplacesDemon}} likes to run its MCMC sampling on a simple number line. Thus, the likelihood
#' function etc. should transform the numbers into the range desired, e.g. 0-1.
#'
#' This function transforms the \code{\link[LaplacesDemon]{LaplacesDemon}} output
#'
#' @param Fit The \code{\link[LaplacesDemon]{LaplacesDemon}} output object.
#' @param MyData The \code{\link[LaplacesDemon]{LaplacesDemon}} input data.
#' @param transfun The function to use for the transformation.
#' @return \code{Fit} The transformed MCMC output.
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite LaplacesDemon_Tutorial
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
reparam_LapDem_output <- function(Fit, MyData, transfun=transform_with_logistic)
{
require(LaplacesDemon)
# Safety
origFit = Fit
# Get the data
BioGeoBEARS_model_object = MyData$BioGeoBEARS_model_object
params_table = BioGeoBEARS_model_object@params_table
freeTF = params_table$type == "free"
for (i in 1:ncol(Fit$Posterior1))
{
minval = params_table$min[freeTF][i]
maxval = params_table$max[freeTF][i]
#######################################################
# Posterior1
#######################################################
# Mark the 0.0s
zerosTF = Fit$Posterior1[, i] == 0.0
# Change the rest to regularspace
Fit$Posterior1[, i] = transfun(paramval=Fit$Posterior1[, i], minval, maxval)
# Convert 0.0s back to 0.0
Fit$Posterior1[zerosTF, i] = 0.0
#######################################################
# Posterior2
#######################################################
if ( !is.null(nrow(Fit$Posterior2)) && (nrow(Fit$Posterior2) >= 1))
{
# Mark the 0.0s
zerosTF = Fit$Posterior2[, i] == 0.0
# Change the rest to regularspace
Fit$Posterior2[, i] = transfun(paramval=Fit$Posterior2[, i], minval, maxval)
# Convert 0.0s back to 0.0
Fit$Posterior2[zerosTF, i] = 0.0
} else {
# If it's 1 row...
for (j in 1:length(Fit$Posterior2))
{
zerosTF = Fit$Posterior2[i] == 0.0
# Change the rest to regularspace
Fit$Posterior2[i] = transfun(paramval=Fit$Posterior2[i], minval, maxval)
# Convert 0.0s back to 0.0
if (zerosTF == TRUE)
{
Fit$Posterior2[i] = 0.0
}
}
}
}
#######################################################
# Re-do the summary, from the parameter-space posterior sample
#######################################################
#######################################################
# re-do the summary, copying from LaplacesDemon()
#######################################################
# Posterior1 is the thinned samples, including everything
LIV = length(Fit$Initial.Values)
thinned = Fit$Posterior1
thinned.rows <- nrow(thinned)
Thinning = Fit$Thinning
Dev <- matrix(Fit$Deviance, ncol=1)
Mon <- Fit$Monitor
Num.Mon <- ncol(Mon)
thinned2 = Fit$Posterior2
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
HD <- BMK.Diagnostic(thinned2, batches = 10)
Ind <- 1 * (HD > 0.5)
BurnIn <- thinned.rows
batch.list <- seq(from = 1, to = nrow(thinned2), by = floor(nrow(thinned2)/10))
for (i in 1:9)
{
if (sum(Ind[, i:9]) == 0)
{
BurnIn <- batch.list[i] - 1
break
}
}
Stat.at <- BurnIn + 1
#rm(batch.list, HD, Ind, thinned2)
}
cat("Assessing Thinning and ESS\n")
acf.temp <- matrix(1, trunc(10 * log10(thinned.rows)), LIV)
ESS1 <- Rec.Thin <- rep(1, LIV)
for (j in 1:LIV) {
temp0 <- acf(thinned[, j], lag.max = nrow(acf.temp),
plot = FALSE)
acf.temp[, j] <- abs(temp0$acf[2:{
nrow(acf.temp) + 1
}, , 1])
ESS1[j] <- ESS(thinned[, j])
Rec.Thin[j] <- which(acf.temp[, j] <= 0.1)[1] * Thinning
}
Rec.Thin <- ifelse(is.na(Rec.Thin), nrow(acf.temp), Rec.Thin)
ESS2 <- ESS(Dev)
ESS3 <- ESS(Mon)
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
if (Stat.at < thinned.rows)
{
ESS4 <- ESS(thinned[Stat.at:thinned.rows, ])
ESS5 <- ESS(Dev[Stat.at:thinned.rows, ])
ESS6 <- ESS(Mon[Stat.at:thinned.rows, ])
}
}
cat("Creating Summaries\n")
Num.Mon <- ncol(Mon)
Summ1 <- matrix(NA, LIV, 7, dimnames = list(MyData$parm.names,
c("Mean", "SD", "MCSE", "ESS", "LB", "Median", "UB")))
Summ1[, 1] <- colMeans(thinned)
Summ1[, 2] <- apply(thinned, 2, sd)
Summ1[, 3] <- 0
Summ1[, 4] <- ESS1
Summ1[, 5] <- apply(thinned, 2, quantile, c(0.025), na.rm = TRUE)
Summ1[, 6] <- apply(thinned, 2, quantile, c(0.5), na.rm = TRUE)
Summ1[, 7] <- apply(thinned, 2, quantile, c(0.975), na.rm = TRUE)
for (i in 1:ncol(thinned)) {
temp <- try(MCSE(thinned[, i]), silent = TRUE)
if (!inherits(temp, "try-error"))
Summ1[i, 3] <- temp
else Summ1[i, 3] <- MCSE(thinned[, i], method = "sample.variance")
}
Deviance <- rep(NA, 7)
Deviance[1] <- mean(Dev)
Deviance[2] <- sd(as.vector(Dev))
temp <- try(MCSE(as.vector(Dev)), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(as.vector(Dev), method = "sample.variance")
Deviance[3] <- temp
Deviance[4] <- ESS2
Deviance[5] <- as.numeric(quantile(Dev, probs = 0.025, na.rm = TRUE))
Deviance[6] <- as.numeric(quantile(Dev, probs = 0.5, na.rm = TRUE))
Deviance[7] <- as.numeric(quantile(Dev, probs = 0.975, na.rm = TRUE))
Summ1 <- rbind(Summ1, Deviance)
for (j in 1:Num.Mon) {
Monitor <- rep(NA, 7)
Monitor[1] <- mean(Mon[, j])
Monitor[2] <- sd(as.vector(Mon[, j]))
temp <- try(MCSE(as.vector(Mon[, j])), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(Mon[, j], method = "sample.variance")
Monitor[3] <- temp
Monitor[4] <- ESS3[j]
Monitor[5] <- as.numeric(quantile(Mon[, j], probs = 0.025,
na.rm = TRUE))
Monitor[6] <- as.numeric(quantile(Mon[, j], probs = 0.5,
na.rm = TRUE))
Monitor[7] <- as.numeric(quantile(Mon[, j], probs = 0.975,
na.rm = TRUE))
Summ1 <- rbind(Summ1, Monitor)
rownames(Summ1)[nrow(Summ1)] <- MyData$mon.names[j]
}
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
Summ2 <- matrix(NA, LIV, 7, dimnames = list(MyData$parm.names,
c("Mean", "SD", "MCSE", "ESS", "LB", "Median", "UB")))
if (Stat.at < thinned.rows)
{
thinned2 <- matrix(thinned[Stat.at:thinned.rows, ], thinned.rows -
Stat.at + 1, ncol(thinned))
Dev2 <- matrix(Dev[Stat.at:thinned.rows, ], thinned.rows -
Stat.at + 1, ncol(Dev))
Mon2 <- matrix(Mon[Stat.at:thinned.rows, ], thinned.rows -
Stat.at + 1, ncol(Mon))
Summ2[, 1] <- colMeans(thinned2)
Summ2[, 2] <- apply(thinned2, 2, sd)
Summ2[, 3] <- 0
Summ2[, 4] <- ESS4
Summ2[, 5] <- apply(thinned2, 2, quantile, c(0.025),
na.rm = TRUE)
Summ2[, 6] <- apply(thinned2, 2, quantile, c(0.5), na.rm = TRUE)
Summ2[, 7] <- apply(thinned2, 2, quantile, c(0.975),
na.rm = TRUE)
for (i in 1:ncol(thinned2))
{
temp <- try(MCSE(thinned2[, i]), silent = TRUE)
if (!inherits(temp, "try-error"))
Summ2[i, 3] <- temp
else Summ2[i, 3] <- MCSE(thinned2[, i], method = "sample.variance")
}
Deviance <- rep(NA, 7)
Deviance[1] <- mean(Dev2)
Deviance[2] <- sd(as.vector(Dev2))
temp <- try(MCSE(as.vector(Dev2)), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(as.vector(Dev2), method = "sample.variance")
Deviance[3] <- temp
Deviance[4] <- ESS5
Deviance[5] <- as.numeric(quantile(Dev2, probs = 0.025,
na.rm = TRUE))
Deviance[6] <- as.numeric(quantile(Dev2, probs = 0.5,
na.rm = TRUE))
Deviance[7] <- as.numeric(quantile(Dev2, probs = 0.975,
na.rm = TRUE))
Summ2 <- rbind(Summ2, Deviance)
for (j in 1:Num.Mon)
{
Monitor <- rep(NA, 7)
Monitor[1] <- mean(Mon2[, j])
Monitor[2] <- sd(as.vector(Mon2[, j]))
temp <- try(MCSE(as.vector(Mon[, j])), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(as.vector(Mon[, j]), method = "sample.variance")
Monitor[3] <- temp
Monitor[4] <- ESS6[j]
Monitor[5] <- as.numeric(quantile(Mon2[, j], probs = 0.025,
na.rm = TRUE))
Monitor[6] <- as.numeric(quantile(Mon2[, j], probs = 0.5,
na.rm = TRUE))
Monitor[7] <- as.numeric(quantile(Mon2[, j], probs = 0.975,
na.rm = TRUE))
Summ2 <- rbind(Summ2, Monitor)
rownames(Summ2)[nrow(Summ2)] <- MyData$mon.names[j]
}
}
Fit$Summary2 = Summ2
}
Fit$Summary1 = Summ1
# Update Deviance, DIC1, DIC2 (actually shouldn't change)
Fit$Deviance = as.vector(Dev)
Fit$DIC1 = c(mean(as.vector(Dev)), var(as.vector(Dev))/2,
mean(as.vector(Dev)) + var(as.vector(Dev))/2)
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
if (Stat.at < thinned.rows)
{
Fit$DIC2 = c(mean(as.vector(Dev2)), var(as.vector(Dev2))/2, mean(as.vector(Dev2)) + var(as.vector(Dev2))/2)
} else {
Fit$DIC2 = rep(NA, 3)
}
}
#######################################################
# Find out how much of each parameter equals 0.0
#######################################################
zeros_TF = Fit$Posterior1 == 0.0
Posterior1_iszero = colSums(zeros_TF) / nrow(Fit$Posterior1)
Fit$Posterior1_iszero = Posterior1_iszero
if ( !is.null(nrow(Fit$Posterior2)) && (nrow(Fit$Posterior2) >= 1))
{
zeros_TF = Fit$Posterior2 == 0.0
Posterior2_iszero = colSums(zeros_TF) / nrow(Fit$Posterior2)
Fit$Posterior2_iszero = Posterior2_iszero
}
return(Fit)
}
#######################################################
# reparam_LapDem_output
#######################################################
#' Change the burnin fraction when re-parameterizing LaplacesDemon MCMC output
#'
#' \code{\link[LaplacesDemon]{LaplacesDemon}} likes to run its MCMC sampling on a simple number line. Thus, the likelihood
#' function etc. should transform the numbers into the range desired, e.g. 0-1.
#'
#' This function changes the burnin fraction
#'
#' @param Fit The \code{\link[LaplacesDemon]{LaplacesDemon}} output object.
#' @param MyData The \code{\link[LaplacesDemon]{LaplacesDemon}} input data.
#' @param burnfract The new burnin fraction
#' @param newsamps If you want to subset the samples.
#' @return \code{Fit} The modified MCMC output.
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite LaplacesDemon_Tutorial
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
change_burnin <- function(Fit, MyData, burnfract=0.1, newsamps=NULL)
{
require(LaplacesDemon)
# Safety
origFit = Fit
# Get the data
BioGeoBEARS_model_object = MyData$BioGeoBEARS_model_object
params_table = BioGeoBEARS_model_object@params_table
freeTF = params_table$type == "free"
# Cut the number of rows by burnfract
rowstart = round(burnfract * nrow(Fit$Posterior1)) + 1
rowend = nrow(Fit$Posterior1)
# The NULL is NOT transferred to seq successfully
if (!is.null(newsamps))
{
rownums = seq(from=rowstart, to=rowend, length.out=newsamps)
} else {
rownums = seq(from=rowstart, to=rowend)
}
# Cut the tables down
Fit$Posterior1 = Fit$Posterior1[rownums, ]
Fit$Monitor = Fit$Monitor[rownums, ]
Fit$Deviance = Fit$Deviance[rownums]
# Revise summary also
Fit$Thinned.Samples = nrow(Fit$Posterior1)
Fit$Iterations = nrow(Fit$Posterior1)
# Maybe also do Posterior2 (should be unnecessary)
dothis = FALSE
if (dothis && !is.null(nrow(Fit$Posterior2)) && (nrow(Fit$Posterior2) >= 1))
{
# Cut the number of rows by burnfract
rowstart = round(burnfract * nrow(Fit$Posterior2)) + 1
rowend = nrow(Fit$Posterior2)
# The NULL is transferred to seq successfully
rownums = seq(start=rowstart, end=rowend, length.out=newsamps)
Fit$Posterior2 = Fit$Posterior2[rownums, ]
}
#######################################################
# Re-do the summary, from the parameter-space posterior sample
#######################################################
#######################################################
# re-do the summary, copying from LaplacesDemon()
#######################################################
# Posterior1 is the thinned samples, including everything
LIV = length(Fit$Initial.Values)
thinned = Fit$Posterior1
thinned.rows <- nrow(thinned)
Thinning = Fit$Thinning
Dev <- matrix(Fit$Deviance, ncol=1)
Mon <- Fit$Monitor
Num.Mon <- ncol(Mon)
thinned2 = Fit$Posterior2
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
HD <- BMK.Diagnostic(thinned2, batches = 10)
Ind <- 1 * (HD > 0.5)
BurnIn <- thinned.rows
batch.list <- seq(from = 1, to = nrow(thinned2), by = floor(nrow(thinned2)/10))
for (i in 1:9)
{
if (sum(Ind[, i:9]) == 0)
{
BurnIn <- batch.list[i] - 1
break
}
}
Stat.at <- BurnIn + 1
#rm(batch.list, HD, Ind, thinned2)
}
cat("Assessing Thinning and ESS\n")
acf.temp <- matrix(1, trunc(10 * log10(thinned.rows)), LIV)
ESS1 <- Rec.Thin <- rep(1, LIV)
for (j in 1:LIV) {
temp0 <- acf(thinned[, j], lag.max = nrow(acf.temp),
plot = FALSE)
acf.temp[, j] <- abs(temp0$acf[2:{
nrow(acf.temp) + 1
}, , 1])
ESS1[j] <- ESS(thinned[, j])
Rec.Thin[j] <- which(acf.temp[, j] <= 0.1)[1] * Thinning
}
Rec.Thin <- ifelse(is.na(Rec.Thin), nrow(acf.temp), Rec.Thin)
ESS2 <- ESS(Dev)
ESS3 <- ESS(Mon)
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
if (Stat.at < thinned.rows)
{
ESS4 <- ESS(thinned[Stat.at:thinned.rows, ])
ESS5 <- ESS(Dev[Stat.at:thinned.rows, ])
ESS6 <- ESS(Mon[Stat.at:thinned.rows, ])
}
}
cat("Creating Summaries\n")
Num.Mon <- ncol(Mon)
Summ1 <- matrix(NA, LIV, 7, dimnames = list(MyData$parm.names,
c("Mean", "SD", "MCSE", "ESS", "LB", "Median", "UB")))
Summ1[, 1] <- colMeans(thinned)
Summ1[, 2] <- apply(thinned, 2, sd)
Summ1[, 3] <- 0
Summ1[, 4] <- ESS1
Summ1[, 5] <- apply(thinned, 2, quantile, c(0.025), na.rm = TRUE)
Summ1[, 6] <- apply(thinned, 2, quantile, c(0.5), na.rm = TRUE)
Summ1[, 7] <- apply(thinned, 2, quantile, c(0.975), na.rm = TRUE)
for (i in 1:ncol(thinned)) {
temp <- try(MCSE(thinned[, i]), silent = TRUE)
if (!inherits(temp, "try-error"))
Summ1[i, 3] <- temp
else Summ1[i, 3] <- MCSE(thinned[, i], method = "sample.variance")
}
Deviance <- rep(NA, 7)
Deviance[1] <- mean(Dev)
Deviance[2] <- sd(as.vector(Dev))
temp <- try(MCSE(as.vector(Dev)), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(as.vector(Dev), method = "sample.variance")
Deviance[3] <- temp
Deviance[4] <- ESS2
Deviance[5] <- as.numeric(quantile(Dev, probs = 0.025, na.rm = TRUE))
Deviance[6] <- as.numeric(quantile(Dev, probs = 0.5, na.rm = TRUE))
Deviance[7] <- as.numeric(quantile(Dev, probs = 0.975, na.rm = TRUE))
Summ1 <- rbind(Summ1, Deviance)
for (j in 1:Num.Mon) {
Monitor <- rep(NA, 7)
Monitor[1] <- mean(Mon[, j])
Monitor[2] <- sd(as.vector(Mon[, j]))
temp <- try(MCSE(as.vector(Mon[, j])), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(Mon[, j], method = "sample.variance")
Monitor[3] <- temp
Monitor[4] <- ESS3[j]
Monitor[5] <- as.numeric(quantile(Mon[, j], probs = 0.025,
na.rm = TRUE))
Monitor[6] <- as.numeric(quantile(Mon[, j], probs = 0.5,
na.rm = TRUE))
Monitor[7] <- as.numeric(quantile(Mon[, j], probs = 0.975,
na.rm = TRUE))
Summ1 <- rbind(Summ1, Monitor)
rownames(Summ1)[nrow(Summ1)] <- MyData$mon.names[j]
}
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
Summ2 <- matrix(NA, LIV, 7, dimnames = list(MyData$parm.names,
c("Mean", "SD", "MCSE", "ESS", "LB", "Median", "UB")))
if (Stat.at < thinned.rows)
{
thinned2 <- matrix(thinned[Stat.at:thinned.rows, ], thinned.rows -
Stat.at + 1, ncol(thinned))
Dev2 <- matrix(Dev[Stat.at:thinned.rows, ], thinned.rows -
Stat.at + 1, ncol(Dev))
Mon2 <- matrix(Mon[Stat.at:thinned.rows, ], thinned.rows -
Stat.at + 1, ncol(Mon))
Summ2[, 1] <- colMeans(thinned2)
Summ2[, 2] <- apply(thinned2, 2, sd)
Summ2[, 3] <- 0
Summ2[, 4] <- ESS4
Summ2[, 5] <- apply(thinned2, 2, quantile, c(0.025),
na.rm = TRUE)
Summ2[, 6] <- apply(thinned2, 2, quantile, c(0.5), na.rm = TRUE)
Summ2[, 7] <- apply(thinned2, 2, quantile, c(0.975),
na.rm = TRUE)
for (i in 1:ncol(thinned2))
{
temp <- try(MCSE(thinned2[, i]), silent = TRUE)
if (!inherits(temp, "try-error"))
Summ2[i, 3] <- temp
else Summ2[i, 3] <- MCSE(thinned2[, i], method = "sample.variance")
}
Deviance <- rep(NA, 7)
Deviance[1] <- mean(Dev2)
Deviance[2] <- sd(as.vector(Dev2))
temp <- try(MCSE(as.vector(Dev2)), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(as.vector(Dev2), method = "sample.variance")
Deviance[3] <- temp
Deviance[4] <- ESS5
Deviance[5] <- as.numeric(quantile(Dev2, probs = 0.025,
na.rm = TRUE))
Deviance[6] <- as.numeric(quantile(Dev2, probs = 0.5,
na.rm = TRUE))
Deviance[7] <- as.numeric(quantile(Dev2, probs = 0.975,
na.rm = TRUE))
Summ2 <- rbind(Summ2, Deviance)
for (j in 1:Num.Mon)
{
Monitor <- rep(NA, 7)
Monitor[1] <- mean(Mon2[, j])
Monitor[2] <- sd(as.vector(Mon2[, j]))
temp <- try(MCSE(as.vector(Mon[, j])), silent = TRUE)
if (inherits(temp, "try-error"))
temp <- MCSE(as.vector(Mon[, j]), method = "sample.variance")
Monitor[3] <- temp
Monitor[4] <- ESS6[j]
Monitor[5] <- as.numeric(quantile(Mon2[, j], probs = 0.025,
na.rm = TRUE))
Monitor[6] <- as.numeric(quantile(Mon2[, j], probs = 0.5,
na.rm = TRUE))
Monitor[7] <- as.numeric(quantile(Mon2[, j], probs = 0.975,
na.rm = TRUE))
Summ2 <- rbind(Summ2, Monitor)
rownames(Summ2)[nrow(Summ2)] <- MyData$mon.names[j]
}
}
Fit$Summary2 = Summ2
}
Fit$Summary1 = Summ1
# Update Deviance, DIC1, DIC2 (actually shouldn't change)
Fit$Deviance = as.vector(Dev)
Fit$DIC1 = c(mean(as.vector(Dev)), var(as.vector(Dev))/2,
mean(as.vector(Dev)) + var(as.vector(Dev))/2)
if ( !is.null(nrow(thinned2)) && (nrow(thinned2) >= 1))
{
if (Stat.at < thinned.rows)
{
Fit$DIC2 = c(mean(as.vector(Dev2)), var(as.vector(Dev2))/2, mean(as.vector(Dev2)) + var(as.vector(Dev2))/2)
} else {
Fit$DIC2 = rep(NA, 3)
}
}
#######################################################
# Find out how much of each parameter equals 0.0
#######################################################
zeros_TF = Fit$Posterior1 == 0.0
Posterior1_iszero = colSums(zeros_TF) / nrow(Fit$Posterior1)
Fit$Posterior1_iszero = Posterior1_iszero
if ( !is.null(nrow(Fit$Posterior2)) && (nrow(Fit$Posterior2) >= 1))
{
zeros_TF = Fit$Posterior2 == 0.0
Posterior2_iszero = colSums(zeros_TF) / nrow(Fit$Posterior2)
Fit$Posterior2_iszero = Posterior2_iszero
}
return(Fit)
}
#######################################################
# Bayesian models, for LaplacesDemon
#######################################################
# params = c(0.1, 0.1, 0.3)
# Calculate the log posterior probability
#######################################################
# LapDem_Model_calc_logPP
#######################################################
#' The function calculating posterior probability of parameter values
#'
#' What it says
#'
#' @param params the parameter values
#' @param MyData The \code{\link[LaplacesDemon]{LaplacesDemon}} input data.
#' @return \code{Modelout} The MCMC output.
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite LaplacesDemon_Tutorial
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
LapDem_Model_calc_logPP <- function(params, MyData)
{
require(LaplacesDemon) # for logit, invlogit
orig_params = params
# Get the data
tr = MyData$tr
tipranges = MyData$tipranges
BioGeoBEARS_model_object = MyData$BioGeoBEARS_model_object
params_table = BioGeoBEARS_model_object@params_table
freeTF = params_table$type == "free"
# NO: Convert the params (range -Inf -> +Inf) to the min/max values
# convert anything outside the range, to min/max, using the interval function
params
for (p in 1:length(params))
{
# Check for parameters being set to 0 by RJMCMC
if (params[p] == 0.0)
{
params[p] = 0.0
} else {
# Inverse logit of param
# params[p] = invlogit(params[p])
# Reset things outside the boundary to (min-max)
minval = params_table$min[freeTF][p]
maxval = params_table$max[freeTF][p]
params[p] = transform_with_logistic(paramval=params[p], minval, maxval)
}
}
params
# NO, INTERVAL() CRASHES with mapply or anything else
# d = 0.1
# e = 0.1
# j = 0.3
# params = c(d,e,j)
# #params = c(d,e)
#
# params_table = BioGeoBEARS_model_object@params_table
# freeTF = params_table$type == "free"
#
#
# # Re-parameterize/constrain with LaplacesDemon:::interval()
# params_not_zero_TF = params != 0.0
# minvals = params_table$min[freeTF][params_not_zero_TF]
# maxvals = params_table$min[freeTF][params_not_zero_TF]
# mapply(
# Input the current parameters into the BioGeoBEARS model object
BioGeoBEARS_model_object = params_into_BioGeoBEARS_model_object(BioGeoBEARS_model_object, params)
# Update the dependent parameters accordingly
BioGeoBEARS_model_object = calc_linked_params_BioGeoBEARS_model_object(BioGeoBEARS_model_object)
# Input the basics for calc_loglike_for_optim()
tip_condlikes_of_data_on_each_state = MyData$tip_condlikes_of_data_on_each_state
print_optim = MyData$print_optim
areas_list = MyData$areas_list
states_list = MyData$states_list
force_sparse = MyData$force_sparse
cluster_already_open = MyData$cluster_already_open
### Parameters
params
### Log of Prior Densities (these are all uniform)
#beta.prior <- sum(dnormv(beta, 0, 1000, log=TRUE))
#sigma.prior <- dhalfcauchy(sigma, 25, log=TRUE)
params_table = BioGeoBEARS_model_object@params_table
freeTF = params_table$type == "free"
rownums = (1:nrow(params_table))[freeTF]
LnPrior = 0
for (rn in rownums)
{
# Assume a Unif(min, max) prior for all variables
LnPrior = LnPrior + dunif(x=params_table$est[rn], min=params_table$min[rn], max=params_table$max[rn], log=TRUE)
# Assume a very flat, but normal, prior, for all variables
# (This will avoid -- NAH, doesn't work cause variables are in regular flat space, not logit space
# NO DOESNT WORK
#LnPrior = LnPrior + rnorm(x=params_table$est[rn], min=params_table$min[rn], max=params_table$max[rn], log=TRUE)
# Try a beta prior with shape1 & shape2 close to 1.00001
plotthis = '
x = seq(0,1,0.01)
y=dbeta(x, shape1=1.01, shape2=1.01)
plot(x,y)
dbeta(x, shape1=1.01, shape2=1.01, log=TRUE)
'
# if ( (rownames(params_table)[rn] == "d") || (rownames(params_table)[rn] == "e"))
# {
# # beta prior for d and e
# LnPrior = LnPrior + dbeta(x=params_table$est[rn], shape1=1.000001, shape2=1.000001, log=TRUE)
# } else {
# # unif prior for j etc.
# LnPrior = LnPrior + dunif(x=params_table$est[rn], min=params_table$min[rn], max=params_table$max[rn], log=TRUE)
# }
#
}
### Log-Likelihood
# mus are the coefficients of the betas
# LL is the sum of the logs of the deviance/densities under the model
#mu <- tcrossprod(Data$X, t(beta))
#LL <- sum(dnorm(Data$y, mu, sigma, log=TRUE))
LnL = calc_loglike_for_optim(params, BioGeoBEARS_model_object, phy=tr, tip_condlikes_of_data_on_each_state, print_optim, areas_list=areas_list, states_list=states_list, force_sparse=force_sparse, cluster_already_open=cluster_already_open)
### Log-Posterior
LnP <- LnL + LnPrior
### yhat = data simulated under these model parameters, for posterior predictive sampling/testing
### (I think)
# yhat = rnorm(length(mu), mu, sigma)
yhat = NA
# Convert the params back to (range -Inf -> +Inf) to the min/max values
# for (p in 1:length(params))
# {
# # Scale to logit
# minval = params_table$min[freeTF][p]
# maxval = params_table$max[freeTF][p]
#
#
# if (params[p] == 0.0)
# {
# # Convert to -70 in logit space, to keep parameter "off the bottom" of precision
# params[p] = -70.0
# } else {
# # Don't subtract minval, as that results in 0, which results in -Inf under logit
# params[p] = (params[p] - minval) / (maxval-minval)
#
# # logit of param
# params[p] = logit(params[p])
# }
# }
### Model output to return
Modelout <- list(LP=LnP, Dev=-2*LnL, Monitor=c(LnP,LnL,LnPrior), yhat=yhat, parm=orig_params)
return(Modelout)
}
# Calculate the log posterior probability
# This version attempts rescaling rather than re-writing parameters...
#######################################################
# LapDem_Model_calc_logPP_orig
#######################################################
#' The function calculating posterior probability of parameter values
#'
#' What it says
#'
#' @param params the parameter values
#' @param MyData The \code{\link[LaplacesDemon]{LaplacesDemon}} input data.
#' @return \code{Modelout} The MCMC output.
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite LaplacesDemon_Tutorial
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
LapDem_Model_calc_logPP_orig <- function(params, MyData)
{
require(LaplacesDemon) # for logit, invlogit
# Get the data
tr = MyData$tr
tipranges = MyData$tipranges
BioGeoBEARS_model_object = MyData$BioGeoBEARS_model_object
params_table = BioGeoBEARS_model_object@params_table
freeTF = params_table$type == "free"
# Convert the params (range -Inf -> +Inf) to the min/max values
for (p in 1:length(params))
{
# Check for parameters being set to 0 by RJMCMC
if (params[p] == 0.0)
{
params[p] = 0.0
} else {
# Inverse logit of param
params[p] = invlogit(params[p])
# Scale to (min-max)
minval = params_table$min[freeTF][p]
maxval = params_table$max[freeTF][p]
params[p] = minval + (maxval-minval) * params[p]
}
}
# Input the current parameters into the BioGeoBEARS model object
BioGeoBEARS_model_object = params_into_BioGeoBEARS_model_object(BioGeoBEARS_model_object, params)
# Update the dependent parameters accordingly
BioGeoBEARS_model_object = calc_linked_params_BioGeoBEARS_model_object(BioGeoBEARS_model_object)
# Input the basics for calc_loglike_for_optim()
tip_condlikes_of_data_on_each_state = MyData$tip_condlikes_of_data_on_each_state
print_optim = MyData$print_optim
areas_list = MyData$areas_list
states_list = MyData$states_list
force_sparse = MyData$force_sparse
cluster_already_open = MyData$cluster_already_open
### Parameters
params
### Log of Prior Densities (these are all uniform)
#beta.prior <- sum(dnormv(beta, 0, 1000, log=TRUE))
#sigma.prior <- dhalfcauchy(sigma, 25, log=TRUE)
params_table = BioGeoBEARS_model_object@params_table
freeTF = params_table$type == "free"
rownums = (1:nrow(params_table))[freeTF]
LnPrior = 0
for (rn in rownums)
{
# Assume a Unif(min, max) prior for all variables
#LnPrior = LnPrior + dunif(x=params_table$est[rn], min=params_table$min[rn], max=params_table$max[rn], log=TRUE)
# Assume a very flat, but normal, prior, for all variables
# (This will avoid -- NAH, doesn't work cause variables are in regular flat space, not logit space
# NO DOESNT WORK
#LnPrior = LnPrior + rnorm(x=params_table$est[rn], min=params_table$min[rn], max=params_table$max[rn], log=TRUE)
# Try a beta prior with shape1 & shape2 close to 1.00001
plotthis = '
x = seq(0,1,0.01)
y=dbeta(x, shape1=1.01, shape2=1.01)
plot(x,y)
dbeta(x, shape1=1.01, shape2=1.01, log=TRUE)
'
if ( (rownames(params_table)[rn] == "d") || (rownames(params_table)[rn] == "e"))
{
# beta prior for d and e
LnPrior = LnPrior + dbeta(x=params_table$est[rn], shape1=1.000001, shape2=1.000001, log=TRUE)
} else {
# unif prior for j etc.
LnPrior = LnPrior + dunif(x=params_table$est[rn], min=params_table$min[rn], max=params_table$max[rn], log=TRUE)
}
}
### Log-Likelihood
# mus are the coefficients of the betas
# LL is the sum of the logs of the deviance/densities under the model
#mu <- tcrossprod(Data$X, t(beta))
#LL <- sum(dnorm(Data$y, mu, sigma, log=TRUE))
LnL = calc_loglike_for_optim(params, BioGeoBEARS_model_object, phy=tr, tip_condlikes_of_data_on_each_state, print_optim, areas_list=areas_list, states_list=states_list, force_sparse=force_sparse, cluster_already_open=cluster_already_open)
### Log-Posterior
LnP <- LnL + LnPrior
### yhat = data simulated under these model parameters, for posterior predictive sampling/testing
### (I think)
# yhat = rnorm(length(mu), mu, sigma)
yhat = NA
# Convert the params back to (range -Inf -> +Inf) to the min/max values
for (p in 1:length(params))
{
# Scale to logit
minval = params_table$min[freeTF][p]
maxval = params_table$max[freeTF][p]
if (params[p] == 0.0)
{
# Convert to -70 in logit space, to keep parameter "off the bottom" of precision
params[p] = -70.0
} else {
# Don't subtract minval, as that results in 0, which results in -Inf under logit
params[p] = (params[p] - minval) / (maxval-minval)
# logit of param
params[p] = logit(params[p])
}
}
### Model output to return
Modelout <- list(LP=LnP, Dev=-2*LnL, Monitor=c(LnP,LnL,LnPrior), yhat=yhat, parm=params)
return(Modelout)
}
#######################################################
# PLotting functions
#######################################################
#LaplacesDemon:::plot.demonoid <- function (x, BurnIn = 0, Data = NULL, PDF = FALSE, Parms = NULL,
# ...)
#######################################################
# plot_demonoid2
#######################################################
#' Plot LaplacesDemon MCMC output
#'
#' Modified version of \code{\link[LaplacesDemon]{plot.demonoid}} to plot to PDF.
#'
#' @param x The MCMC output
#' @param BurnIn The burnin amount
#' @param Data The \code{\link[LaplacesDemon]{LaplacesDemon}} input data.
#' @param PDF If TRUE, plot to PDF.
#' @param Parms The parameters.
#' @param ... Additional arguments to standard functions
#' @return Nothing.
#' @export
#' @seealso \code{\link[LaplacesDemon]{LaplacesDemon}}
#' @note Go BEARS!
#' @author Nicholas J. Matzke \email{matzke@@berkeley.edu}
#' @references
#' \url{http://phylo.wikidot.com/matzke-2013-international-biogeography-society-poster}
#' @bibliography /Dropbox/_njm/__packages/BioGeoBEARS_setup/BioGeoBEARS_refs.bib
#' @cite LaplacesDemon_Tutorial
#' @cite Matzke_2012_IBS
#' @examples
#' test=1
plot_demonoid2 <- function (x, BurnIn = 0, Data = NULL, PDF = FALSE, Parms = NULL,
...)
{
greyval1 = 0.75
greyval1 = rgb(greyval1, greyval1, greyval1)
greyval2 = 0.35
greyval2 = rgb(greyval2, greyval2, greyval2)
if (missing(x))
stop("The x argument is required.")
if (class(x) != "demonoid")
stop("x must be of class demonoid.")
if (is.null(Data))
stop("The Data argument is NULL.")
if (BurnIn >= nrow(x$Posterior1))
BurnIn <- 0
Stat.at <- BurnIn + 1
if (is.null(Parms)) {
Posterior <- x$Posterior1
}
else {
Parms <- sub("\\[", "\\\\[", Parms)
Parms <- sub("\\]", "\\\\]", Parms)
Parms <- sub("\\.", "\\\\.", Parms)
if (length(grep(Parms[1], colnames(x$Posterior1))) ==
0)
stop("Parameter in Parms does not exist.")
keepcols <- grep(Parms[1], colnames(x$Posterior1))
if (length(Parms) > 1) {
for (i in 2:length(Parms)) {
if (length(grep(Parms[i], colnames(x$Posterior1))) ==
0)
stop("Parameter in Parms does not exist.")
keepcols <- c(keepcols, grep(Parms[i], colnames(x$Posterior1)))
}
}
Posterior <- as.matrix(x$Posterior1[, keepcols])
colnames(Posterior) <- colnames(x$Posterior1)[keepcols]
}
if (PDF == TRUE) {
pdf("LaplacesDemon.Plots.pdf")
par(mfrow = c(3, 3))
}
else {
par(mfrow = c(3, 3), ask = TRUE)
}
for (j in 1:ncol(Posterior)) {
plot(Stat.at:x$Thinned.Samples, Posterior[Stat.at:x$Thinned.Samples,
j], type = "l", xlab = "Generations", ylab = "Value",
main = colnames(Posterior)[j])
panel.smooth(Stat.at:x$Thinned.Samples, Posterior[Stat.at:x$Thinned.Samples,
j], pch = "")
###########################################
# Plot the parameter values
###########################################
numbreaks = 30
vals = pretty(Posterior[Stat.at:x$Thinned.Samples, j])
breakpoints = seq(from=min(vals), to=max(vals), length.out=numbreaks)
# Get values for density plot
histres2 = hist(x=Posterior[Stat.at:x$Thinned.Samples, j], breaks=100, plot=FALSE)
histres = hist(x=Posterior[Stat.at:x$Thinned.Samples, j], breaks=breakpoints, freq=FALSE, add=FALSE, xlab="Value", ylab="Density",
col=greyval1, border=greyval2, main=colnames(Posterior)[j])
lines(density(unlist(mapply(rep,histres2$mids,histres2$counts)), adjust=1), col="black", lwd=1.5)
#polygon(density(Posterior[Stat.at:x$Thinned.Samples, j]), col = "white", border = "black")
abline(v = 0, col = "red", lty = 2)
if (!is.constant(Posterior[Stat.at:x$Thinned.Samples,
j])) {
z <- acf(Posterior[Stat.at:x$Thinned.Samples, j],
plot = FALSE)
se <- 1/sqrt(length(Posterior[Stat.at:x$Thinned.Samples,
j]))
plot(z$lag, z$acf, ylim = c(min(z$acf, -2 * se),
1), type = "h", main = colnames(Posterior)[j],
xlab = "Lag", ylab = "Correlation")
abline(h = (2 * se), col = "red", lty = 2)
abline(h = (-2 * se), col = "red", lty = 2)
}
else {
plot(0, 0, main = paste(colnames(Posterior)[j], "is a constant."))
}
}
plot(Stat.at:length(x$Deviance), x$Deviance[Stat.at:length(x$Deviance)],
type = "l", xlab = "Iterations", ylab = "Value", main = "Deviance")
panel.smooth(Stat.at:length(x$Deviance), x$Deviance[Stat.at:length(x$Deviance)],
pch = "")
#plot(density(x$Deviance[Stat.at:length(x$Deviance)]), xlab = "Value",
# main = "Deviance")
#polygon(density(x$Deviance[Stat.at:length(x$Deviance)]),
# col = "black", border = "black")
# Get values for density plot
###########################################
# Plot the Deviance values
###########################################
numbreaks = 30
vals = pretty(x$Deviance[Stat.at:length(x$Deviance)])
breakpoints = seq(from=min(vals), to=max(vals), length.out=numbreaks)
histres2 = hist(x=x$Deviance[Stat.at:length(x$Deviance)], breaks=100, plot=FALSE)
histres = hist(x=x$Deviance[Stat.at:length(x$Deviance)], breaks=breakpoints, freq=FALSE, add=FALSE, xlab="Value", ylab="Density",
col=greyval1, border=greyval2, main="Deviance")
lines(density(unlist(mapply(rep,histres2$mids,histres2$counts)), adjust=1), col="black", lwd=1.5)
abline(v = 0, col = "red", lty = 2)
if (!is.constant(x$Deviance[Stat.at:length(x$Deviance)])) {
z <- acf(x$Deviance[Stat.at:length(x$Deviance)], plot = FALSE)
se <- 1/sqrt(length(x$Deviance[Stat.at:length(x$Deviance)]))
plot(z$lag, z$acf, ylim = c(min(z$acf, -2 * se), 1),
type = "h", main = "Deviance", xlab = "Lag", ylab = "Correlation")
abline(h = (2 * se), col = "red", lty = 2)
abline(h = (-2 * se), col = "red", lty = 2)
}
else {
plot(0, 0, main = "Deviance is a constant.")
}
if (is.vector(x$Monitor)) {
J <- 1
nn <- length(x$Monitor)
}
else if (is.matrix(x$Monitor)) {
J <- ncol(x$Monitor)
nn <- nrow(x$Monitor)
}
for (j in 1:J) {
plot(Stat.at:nn, x$Monitor[Stat.at:nn, j], type = "l",
xlab = "Iterations", ylab = "Value", main = Data$mon.names[j])
panel.smooth(Stat.at:nn, x$Monitor[Stat.at:nn, j], pch = "")
#plot(density(x$Monitor[Stat.at:nn, j]), xlab = "Value",
# main = Data$mon.names[j])
#polygon(density(x$Monitor[Stat.at:nn, j]), col = "black",
# border = "black")
###########################################
# Plot the LP, LnL, LnPrior values
###########################################
numbreaks = 30
vals = pretty(x$Monitor[Stat.at:nn, j])
breakpoints = seq(from=min(vals), to=max(vals), length.out=numbreaks)
histres2 = hist(x=x$Monitor[Stat.at:nn, j], breaks=100, plot=FALSE)
histres = hist(x=x$Monitor[Stat.at:nn, j], breaks=breakpoints, freq=FALSE, add=FALSE, xlab="Value", ylab="Density",
col=greyval1, border=greyval2, main=Data$mon.names[j])
lines(density(unlist(mapply(rep,histres2$mids,histres2$counts)), adjust=1), col="black", lwd=1.5)
abline(v = 0, col = "red", lty = 2)
if (!is.constant(x$Monitor[Stat.at:nn, j])) {
z <- acf(x$Monitor[Stat.at:nn, j], plot = FALSE)
se <- 1/sqrt(length(x$Monitor[Stat.at:nn, j]))
plot(z$lag, z$acf, ylim = c(min(z$acf, -2 * se),
1), type = "h", main = Data$mon.names[j], xlab = "Lag",
ylab = "Correlation")
abline(h = (2 * se), col = "red", lty = 2)
abline(h = (-2 * se), col = "red", lty = 2)
}
else {
plot(0, 0, main = paste(Data$mon.names[j], "is a constant."))
}
}
if (nrow(x$CovarDHis) > 1) {
if (x$Algorithm %in% c("Adaptive Hamiltonian Monte Carlo",
"Hamiltonian Monte Carlo with Dual-Averaging", "No-U-Turn Sampler")) {
plot(x$CovarDHis[, 1], type = "l", xlab = "Adaptations",
ylab = expression(epsilon))
}
else {
Diff <- abs(diff(x$CovarDHis))
adaptchange <- matrix(NA, nrow(Diff), 3)
for (i in 1:nrow(Diff)) {
adaptchange[i, 1:3] <- as.vector(quantile(Diff[i,
], probs = c(0.025, 0.5, 0.975)))
}
plot(adaptchange[, 2], ylim = c(min(adaptchange),
max(adaptchange)), type = "l", col = "red", xlab = "Adaptations",
ylab = "Absolute Difference", main = "Proposal Variance",
sub = "Median=Red, 95% Bounds=Gray")
lines(adaptchange[, 1], col = "gray")
lines(adaptchange[, 3], col = "gray")
}
}
if (PDF == TRUE)
dev.off()
}
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.