plotChains: MCMC Diagnostic Plots

In abnormally-distributed/Bayezilla: Bayesian Models With JAGS and Several Utilities for Data Exploration and Model Evaluation.

Description

Displays traceplots and cumulative means. Draws on Edward Tufte's minimalist aesthetic ideas for the design.

Usage

plotChains(fit, keeppars = NULL, droppars = c("ySim", "log_lik"),
  col = "blues", nrow = 4, ncol = 2)

Arguments

fit	the stanfit or runjags object
keeppars	the names of the parameters to be plotted.
col	the color scheme. One of "blue" (the default), "darkblue", "purples", "reds", "greens", "mix" , or "mix2"
nrow	number of rows in the layout
ncol	number of columns in the layout

Details

Traceplots should look very fuzzy and chaotic. If they look more ordered and correlated that indicates a problem with the sampler or model specification.

Cumulative mean plots display how quickly the model converges onto the expected value. The chains should all converge onto the same value or very close value very quickly. If they do not, they indicates you either need to draw more samples, use a longer adaptation and/or burnin/warmup for the MCMC chains, or perhaps reconsider the model if those do not work.

Value

a base R plot

Examples

plotChains()

abnormally-distributed/Bayezilla documentation built on Oct. 31, 2019, 1:57 a.m.