Skip to contents

Graphical diagnoses to check if the simulation converges

Source: R/check_conv.R
check_conv.Rd

Compare the distribution of different stages of simulation (for plot_type == "bin" or plot_type = "density"), or show how the percentiles of the distribution evolve over time (for plot_type == cumuplot, see coda::cumuplot() for details). More convergence checking methods for MCMC data are available at the coda package. Be cautious: each convergence checking method has its shortcomings, so do not blindly use any results as the definitive conclusion that a simulation converges or not.

Usage

check_conv(output, vars, sample_perc = 0.2, plot_type = "bin")

# S3 method for check_conv
print(x, ask = TRUE, ...)

Arguments

output

A matrix of simulation output.

vars

The names of variables to check.

sample_perc

The percentage of data sample for the initial, middle, and final stage of the simulation. Not required if plot_type == "cumuplot".

plot_type

Which type of plots should be generated? ("bin", "density", or "cumuplot" which uses coda::cumuplot())

x

The object.

ask

Ask to press enter to see the next plot?

...

Not in use.

Value

A check_conv object that contains the convergence checking result(for plot_type == "bin" or plot_type = "density"), or draw the cumuplot without a return value (for plot_type == cumuplot).

Methods (by generic)

  • print(check_conv): Print a check_conv object.