R/pdisc.R

In bayesball/LearnBayes: Learning Bayesian Inference

"pdisc" <-
function(p,prior,data)
{ 
# PDISC Posterior distribution for a proportion with discrete models.
#  POST = PDISC(P,PRIOR,DATA) returns a vector of posterior probabilities.
#  P is the vector of values of the proportion, PRIOR is the corresponding
#  vector of prior probabilities and DATA is the vector of data (number of
#  successes and failures in set of independent Bernoulli trials
#------------------------
# Written by Jim Albert
# albert@bgnet.bgsu.edu
# November 2004
#------------------------

s=data[1]; f=data[2]
p1=p+.5*(p==0)-.5*(p==1)

like=s*log(p1)+f*log(1-p1)
like=like*(p>0)*(p<1)-999*((p==0)*(s>0)+(p==1)*(f>0))
like=exp(like-max(like))

product=like*prior
post=product/sum(product)

return(post)
}