BEST - Bayesian t.test
It is interesting to implement the t.test using mcmc based on paper written by J.K. Kruschke
Implemented using MCMCpack
from the high level, 5 parameters to estimate - mu1, mu2, sigma1, sigma2, nu
prior mu ~ N(mu, sd*1000, log=T)
prior sd ~ unif(sd, sd/1000, sd*1000, log=T)
prior nu ~ dexp(nu-1, 1/29, log=T)
likelihood <- function(theta, x, y)
log.prior + sum(t(mu, sd, nu, log=T))
