calc.mult.GR: Calculate Gelman and Rubin diagnostic multiple times for...

In bbolker/mixstock: Functions for Mixed Stock Analysis

Calculate Gelman and Rubin diagnostic multiple times for mixed stock analyses

Description

Runs Gelman and Rubin diagnostics from CODA multiple times, to get an idea of the variation in convergence statistics

Usage

calc.mult.GR(data, n=10, tot=20000, burn=1, verbose=FALSE)

Arguments

data	Mixed stock analysis data (a list with elements sourcesamp and mixsamp
n	Number of replicates to run
tot	Total number of iterates for each chain
burn	Burn-in time for each chain
verbose	Produce verbose output?

Details

Runs calc.GR multiple times, produces a summary table of the maximum point estimate and maximum 97.5% estimate (across all variables) for each run with a different random-number seed.

Value

A matrix with each row giving the random-number seed, max. point estimate, max. 97.5% quantile for each run.

Note

The generally accepted criteria for declaring convergence according to Gelman and Rubin is that all of the 97.5% quantiles of the estimates of shrink factors are less than 1.2.

Author(s)

Ben Bolker

See Also

calc.GR

Examples

data(simex)
calc.GR(simex,tot=2000)

bbolker/mixstock documentation built on July 23, 2024, 12:18 p.m.