contLikMCMC: contLikMCMC

In euroformix: Package for intepretation of quantitative DNA mixtures

Description

contLikMCMC simulates from the posterior distribution for a bayesian STR DNA mixture model.

Usage

contLikMCMC(mlefit, niter = 10000, delta = 10, maxxi = 1)

Arguments

mlefit	Fitted object using contLikMLE
niter	Number of samples in the MCMC-sampling.
delta	A numerical parameter to scale with the covariance function Sigma. Default is 2. Should be higher to obtain lower acception rate.
uppermu	Upper boundary of the mu-parameters
uppersigma	Upper boundary of the sigma-parameters
upperxi	Upper boundary of the xi-parameters

Details

The procedure are doing MCMC to approximate the marginal probability over noisance parameters. Mixture proportions have flat prior.

If no initial values or covariance matrix has been provided to the function, a call to the MLE function is applied. The Metropolis Hastings routine uses a Multivariate Normal distribution with mean 0 and covariance as delta multiplied with the inverse negative hessian with MLE inserted as transistion kernel. Function calls procedure in c++ by using the package Armadillo and Boost. Marginalized likelihood is estimated using Metropolis Hastings with the "Gelfand and Dey" method.

Value

ret A list (margL,posttheta,postlogL,logpX,accrat,Ubound ) where margL is Marginalized likelihood for hypothesis (model) given observed evidence, posttheta is the posterior samples from a MC routine, postlogL is sampled log-likelihood values, accrat is ratio of accepted samples. Ubound is upper boundary of parameters.

Author(s)

Oyvind Bleka <Oyvind.Bleka.at.fhi.no>

References

Craiu,R.V. and Rosenthal, J.S. (2014). Bayesian Computation Via Markov Chain Monte Carlo. Annu. Rev. Stat. Appl., 1,179-201.

euroformix documentation built on May 2, 2019, 4:48 p.m.