plotchain: Diagnostic plots for HDP posterior sampling chain

In nicolaroberts/hdp: Hierarchical Dirichlet Process for categorical count data

Description

Diagnostic plots for HDP posterior sampling chain

Usage

plot_lik(chain, start = 1, end = length(lik(chain)), col_lik = "blue",
  col_burn = "red", xlab = "Iteration", ylab = "Likelihood", ...)

plot_numcluster(chain, col = "blue", xlab = "Sample",
  ylab = "Number of raw clusters", ...)

plot_data_assigned(chain, legend = TRUE, col_early = "hotpink",
  col_late = "skyblue3", dat_prop = 0.995,
  xlab = "Number of raw clusters",
  ylab = "Cumulative prop. of data assigned", ...)

Arguments

chain	A hdpSampleChain object
start	The starting iteration to plot from (default 1)
end	The final iteration to plot to (default is end of chain)
col_lik	Plot colour of likelihood (default blue)
col_burn	Plot colour of burnin (default red)
xlab	Horizontal axis label
ylab	Vertical axis label
...	Other arguments to plot
col	Plot colour for numcluster (default blue)
legend	Logical - should a legend be included? (default TRUE)
col_early	Color ramp side for early posterior samples
col_late	Color ramp side for late posterior samples
dat_prop	Extend horiztonal axis to dat_prop proportion of data assigned

Examples

par(mfrow=c(2,2))
lapply(chains(mut_example_multi), plot_lik, bty="L", start=1000)
lapply(chains(mut_example_multi), plot_numcluster, bty="L")
lapply(chains(mut_example_multi), plot_data_assigned, bty="L")

nicolaroberts/hdp documentation built on May 23, 2019, 5:09 p.m.