MLCVGauss: Inference and model selection under the assumption of...

In Sunder: Quantification of the effect of geographic versus environmental isolation on genetic differentiation

Description

Inference and model selection for analysis of geographical genetic variation under the assumption of Gaussian distribution of allele counts for bi-allelic loci. Parameter estimation by maximization of the likelihood.

Usage

MLCVGauss(gen, D_G, D_E,theta.max,theta.min,ntrain,nresamp)

Arguments

gen	A matrix with dimensions (n,l) n: number of geographical locations, l: number of loci.
D_G	A matrix of geograpical distances
D_E	A matrix of environmental distances
theta.max	Upper bounds for the vector of parameters in theta. Note that in theta, the parameters are assumed to be in this order: (alpha,beta_G, beta_E, gamma, delta)
theta.min	Lower bounds for the vector of parameters in theta. Note that in theta, the parameters are assumed to be in this order: (alpha,beta_G, beta_E, gamma, delta)
ntrain	Number of sites used for training. An integer smaller than nrow(gen). If ntrain is equal to the number of sampling sites, the function estimates parameters on the whole dataset and does not perform cross-validation.
nresamp	Number of resamplings. An integer larger than 1.

Value

A list with either a component named mod.lik (containing likelihoods on the validation set for the various models compared) or a vector of estimated parameters (if ntrain is equal to the number of sampling sites).

Author(s)

Gilles Guillot

Examples

## Not run: 
nsite <- 200
nloc <- 1000
hap.pop.size <- 100
theta <- c(runif(n=1,.5,10),
           runif(n=1,.01,10),
           runif(n=1,.01,10),
           runif(n=1,.5,1),
           runif(n=1,.01,.1)
           )
mod <- 'G+E' 
dat <- SimSunderData(mod=mod,
                     theta=theta,
                     nsite=nsite,
                     nloc=nloc,
                     hap.pop.size=hap.pop.size,
                     nalM=2,nalm=2, #bi-allelic loci
                     var.par=1,
                     scale.par=3)
gen <- dat$gen[,,1]
D_G <- dat$D_G
D_E <- dat$D_E

res <- MLCVGauss(gen,D_G,D_E,
                 ntrain=nrow(gen)/2,
                 nresamp=3)

which.max(res$mod.lik)

## End(Not run)

Sunder documentation built on May 2, 2019, 2:17 p.m.