gibbsC: Run mixed stock analysis Gibbs sampler, in C
In bbolker/mixstock: Functions for Mixed Stock Analysis

View source: R/mixstock.R View source: R/RCS/turtle.R

gibbsC

R Documentation

Run mixed stock analysis Gibbs sampler, in C

Description

Runs a Gibbs sampler MCMC for mixed stock analysis, calling a routine written in C (for code). Low-level function, called by other functions.

Usage

gibbsC(a=1, startiter, maxiter, data, mixsamp=NULL, sourcesamp=NULL,
startfval=NULL, thin=1, fprior=NULL, outfile=FALSE,
outfn="mixstock-gibbs", randseed=1001, rptiter=-1, debug=FALSE,
contrun=FALSE,contrib.start=NULL,sourcefreq.start=NULL)

Arguments

`a`	Prior strength parameter
`startiter`	Number of iterations to discard (burn-in)
`maxiter`	Total number of chain steps
`data`	A `mixstock.data` object
`mixsamp`	Marker sample from mixed populations
`sourcesamp`	Marker samples from sources
`startfval`	Where to start the chain: 0=
`thin`	thinning factor
`fprior`	Bayesian prior
`outfile`	send data to an output file?
`outfn`	name of output file
`randseed`	random-number seed
`rptiter`	frequency for sending reports to screen
`debug`	debug?
`contrun`	continuation run? set fval and sourcefreq directly
`contrib.start`	vector of starting contributions
`sourcefreq.start`	matrix of starting source freqs

Value

a numeric matrix containing samples from the chain: each row is a vector of estimated contribution frequencies from each rookery

Note

gibbsC calls C code to generate multinomial deviates derived from the randlib.c library (version 1.3: currently located at http://lcrocker.github.io/onejoker/randlib/) written by Barry W. Brown, James Lovato, Kathy Russell, and John Venier, derived in turn from page 559 of: Devroye, Luc, Non-Uniform Random Variate Generation. Springer-Verlag, New York, 1986.