POUMM (version 2.1.7)

plot.summary.POUMM: Plot a summary of a POUMM fit

Description

Plot a summary of a POUMM fit

Usage

# S3 method for summary.POUMM
plot(
  x,
  type = c("MCMC"),
  doPlot = TRUE,
  stat = c("alpha", "theta", "sigma", "sigmae", "g0", "H2tMean"),
  chain = NULL,
  doZoomIn = FALSE,
  zoomInFilter = paste0("(stat %in% c('H2e','H2tMean','H2tInf','H2tMax') & ",
    "(value >= 0 & value <= 4="" 1)="" )="" |",="" "(="" !stat="" %in%="" c('h2e','h2tmean','h2tinf','h2tmax')="" &="" ",="" "(value="" <="median(HPDUpper)" +="" *="" (median(hpdupper)="" -="" median(hpdlower))="" &",="" "value="">= median(HPDLower) - 4 * (median(HPDUpper) - median(HPDLower))))"),
  palette = c("#999999", "#0072B2", "#CC79A7", "#E69F00", "#D55E00", "#56B4E9",
    "#009E73", "#F0E442"),
  prettyNames = TRUE,
  ...
)

Arguments

x

An object of class POUMM.

type

A character indicating the type of plot(s) to be generated. Defaults to "MCMC", resulting in a trace and density plot for the selected statistics (see argument stat). Currently, only 'MCMC' type is supported.

doPlot

Logical indicating whether a plot should be printed on the currently active graphics device or whether only to return a list of plot- objects for further processing. Defaults to TRUE.

stat

A character vector with the names of statistics to be plotted. These should be names from the stats-list (see argument statFunctions). Defaults to c("alpha", "theta", "sigma", "sigmae", "H2tMean", "H2tInf").

chain

A vector of integers indicating the chains to be plotted.

doZoomIn

(type MCMC only) A logical value indicating whether the produced plots should have a limitation on the x-axis according to an expression set in zoomInFilter (see below). Default value is FALSE.

zoomInFilter

A character string which evaluates as logical value. If doZoomIn is set to TRUE, this filter is applied to each point in each MCMC chain and the data-point is filtered out if it evaluates to FALSE. This allows to zoomIn the x-axis of density plots but should be used with caution, since filtering out points from the MCMC-sample can affect the kernel densities. Unfortunately, filtering out values is currently the only way to affect the limits of individual facets in ggplot2. The default value is a complicated expression involving the HPD from all MCMC chains (normally one chain from the prior and 2 chains from the posterior): zoomInFilter = paste0("stat The identifiers in this expression can be any column names found in a summary of a POUMM object.

palette

A vector of colors (can be character strings) corresponding to the different chains (in their order 1 (prior), 2, 3). Defaults to c("#999999", "#0072B2", "#CC79A7", "#E69F00", "#D55E00", "#56B4E9", "#009E73", "#F0E442"), which is a color-blind friendly.

prettyNames

A logical indicating if greek letters and sub/superscripts should be used for the names of columns in the posterior density pairs-plot.

...

Not used; included for compatibility with the generic function plot.

Value

If doPlot==TRUE, the function returns nothing and produces output on the current graphics device as a side-effect. Otherwise, the function returns a list of plot-objects: traceplot and densplot.

Examples

# NOT RUN {
library(POUMM)

set.seed(1)

N <- 1000

# create a random non-ultrametric tree of N tips
tree <- ape::rtree(N)  

# Simulate the evolution of a trait along the tree
z <- rVNodesGivenTreePOUMM(
  tree, g0 = 8, alpha = 1, theta = 4, sigma = 1.2, sigmae = .8)

fit <- POUMM(z[1:N], tree, spec = list(nSamplesMCMC = 4e5))

# Summarize the results from the fit in a table:
summary(fit)

# Create plots for some of the inferred parameters/statistics:
pl <- plot(fit, stat = c("alpha", "theta", "sigma", "sigmae", "H2tMean"), 
           doZoomIn = TRUE, 
           zoomInFilter = paste("!(stat %in% c('alpha', 'sigma', 'sigmae')) |",
                                "(value >= 0 & value <= 8)"),
           doPlot = FALSE)

pl$traceplot
pl$densplot
# }
# NOT RUN {
# }