mhReffects: Metropolis-Hastings algorithm for a one-way normal random...
In mas3321: Bayes normal linear model and basic MCMC

Simulates realisations from the posterior distribution for the population mean and precision components in a one-way normal random effects model with a semi-conjugate prior. The method marginalises over the random effects and uses univariate normal or log normal random walk proposals for the precision components.

1 2	mhReffects(N, initial, intau, innu, priorparam, m, n, ybar, s, show = TRUE, innLogscale = FALSE)

`N`	length of MCMC chain
`initial`	starting values for the algorithm
`intau`	standard deviation of normal random walk innovation for data precision parameter tau
`innu`	standard deviation of normal random walk innovation for random effects precision parameter nu
`priorparam`	prior parameters a,b,c,d,e,f
`m`	number of treatments
`n`	vector containing the number of observations on each treatment
`ybar`	vector containing the mean of observations on each treatment
`s`	vector containing the standard deviation of observations on each treatment
`show`	logical. If true then acceptance rate for the proposals will be given
`innLogscale`	logical. If true then proposals are made on a log scale

data(contamination)
n=tapply(contamination$acc,contamination$keyboard,length)
ybar=tapply(contamination$acc,contamination$keyboard,mean)
s=sqrt(tapply(contamination$acc,contamination$keyboard,var)*(n-1)/n)
mcmcAnalysis(mhReffects(N=100,initial=c(200,2e-5,1),intau=1e-5,innu=7.9,priorparam=c(200,0.1,0.1,0.1,0.1,0.1),m=10,n=n,ybar=ybar,s=s,show=TRUE),rows=3)