gold_mcmcplots: Plots a Variety of Convergence Diagnostic Plots

In reykp/biRd: What the Package Does in One 'Title Case' Line

Description

Plots the convergence plots on the parameters from gold.

Usage

gold_mcmcplots(duos_output, type = "traceplot", plots = "all", burnin = 1)

Arguments

gold_output	The list returned by duos containing the density estimate results.
type	The type of convergent plot to create (see details).
npar	The number of paramters to plot. 3X3 grids of tracepltos are created (DEFAULT is 9).
burnin	The desired burnin to discard from the results. By default, it is 1 so that all iterations are plotted.

Details

Options for type

There are several options on which convergence plots to create.

• "traceplot": Creates trace plots on the parameters (DEFAULT).

• "acf": Creates autocorrelation plots using lag=1000. The burnin should be set to the desired value to discard.

• "rm": Creates running mean plots on the parameters.

If npar is specified to be 9, 9 paramters (as equally spaced as possible) are chosen to plot.

Value

A plot of trace plots, acf plots, or running mean pltos.

Examples

## --------------------------------------------------------------------------------
## Uniform Distribution
## --------------------------------------------------------------------------------

# First run 'duos' on data sampled from a Uniform(0,1) distribution with 50 data points.
y <- runif(50)
duos_unif <- duos(y = y)

# Plot the trace plots of the cut-points on a single graph (trace plot is the default)
duos_mcmcplots(duos_unif)

# Plot the acf plots of the cut-points on separate graphs
duos_mcmcplots(duos_unif, type = "acf", plots = "indiv", burnin = 10000)

## --------------------------------------------------------------------------------
## Beta Distribution
## --------------------------------------------------------------------------------

# First run 'duos' on data sampled from a Beta(0.5,0.5) distribution with 300 data points.
y <- rbeta(300, 0.5, 0.5)
duos_arcsin <- duos(y = y, k = 10, MH_N = 20000)

#Plot the trace plots for the cut-points as individual graphs
#Note: The plots are printed six at a time so multiple panels are printed
duos_mcmcplots(duos_arcsin, plots = "indiv")

#Plot the trace plots for the bin proportions as individual graphs
#Note: The plots are printed six at a time so multiple panels are printed
duos_mcmcplots(duos_arcsin, parameters = "p", plots = "indiv")

## --------------------------------------------------------------------------------
## Bimodal Distribution
## --------------------------------------------------------------------------------

# Sample 150 random uniforms
u <- runif(150)
y <- rep(NA, 150)
# Sampling from the mixture
for(i in 1:150){
  if(u[i]<.3){
   y[i] <- rnorm(1, 0, 1)
  }else {
   y[i] <- rnorm(1, 4, 1)
  }
}
# First run 'duos' on data sampled from a bimodal distribution with 150 data points.
duos_bimodal <- duos(y = y, k = 8, MH_N = 20000)

# Plot the running mean plots for the cut-point parameters 
duos_mcmcplots(duos_bimodal, type = "rm", parameters = "c")

# Plot the autocorrelation plots for the bin proportions with a burnin of 10,000
duos_mcmcplots(duos_bimodal, type = "acf", parameters = "p", burnin = 10000)

reykp/biRd documentation built on May 17, 2019, 8:16 p.m.