siarsolomcmcv4: MCMC for stable isotope data with only single target organisms

Description

Runs an MCMC on stable isotope data from certain organisms to determine their dietary habits. This version requires only a single target organism per group

Usage

siarsolomcmcv4(data, sources, corrections = 0, concdep = 0, iterations=200000, burnin=50000, howmany=10000, thinby=15, prior = rep(1, nrow(sources)), siardata=list(SHOULDRUN=FALSE))

Arguments

data

A matrix with each food source as a seperate row and each isotope as a seperate column.

sources

A matrix containing the mean and standard deviations of the fractionated correction values for each of the isotopes. Also allows corrections = 0 for pre-corrected data.

corrections

A matrix containing the mean and standard deviations of the fractional correction values for each of the isotopes. Also allows corrections = 0 for pre-corrected data.

concdep

A matrix containing the mean and standard deviations of the concentration dependence values for each of the isotopes. Also allows concdep = 0 for data with no required concentration dependence. Note that version 4.0 does not use the standard deviations.

iterations

The number of iterations to run.

burnin

The size of the burnin

howmany

How often to report the number of iterations.

thinby

The amount of thinning of the iterations.

prior

The dirichlet distribution prior parameters, the default is rep(1,numsources). New parameters can be estimated via the function siarelicit.

siardata

A list containing some or all of the following parts: targets, sources, corrections, PATH, TITLE, numgroups, numdata, numsources, numiso, SHOULDRUN, GRAPHSONLY, EXIT, and output. For more details of these inputs see the siarmenu function.

Value

A parameter matrix consisting of (iterations-burnin)/thinby rows with numgroups*(numsources+numiso) columns, where numsources is the number of food sources, numiso is the number of isotopes, and numgroups is the number of groups. The parameter matrix is structured so that, for each group, the first columns are those of the proportions of each food source eaten, the next columns are the standard deviations for each isotope. This format repeats across rows to each group. The parameters may then subsequently be used for plotting, convergence checks, summaries, etc, etc.

Details

The model assumes that each target value comes from a Gaussian distribution with an unknown mean and standard deviation. The structure of the mean is a weighted combination of the food sources' isotopic values. The weights are made up dietary proportions (which are given a Dirichlet prior distribution) and the concentration depdendencies given for the different food sources. The standard deviation is divided up between the uncertainty around the fractionation corrections (if corrections are given) and the natural variability between target individuals within a defined group (or between all individuals if no grouping structure is specified). The default iterations numbers work well for the demo data sets, but advanced users will want to adjust them to suit their analysis. Note that this version is analagous to the Moore and Semmens (2008) MixSIR model except with a Dirichlet prior distribution.

References