credint {emdbook}

Calculate Bayesian credible intervals

Description

Calculate Bayesian credible intervals based on various types of information about the posterior distribution

Usage

tcredint(dist, parlist, ranges, level = 0.95, eps = 1e-05,verbose=FALSE)
ncredint(pvec,npost,level=0.95,tol=0.01,verbose=FALSE)

Arguments

dist

character string giving the name of a distribution for which "d", "q", and "p" function exist, e.g. "beta"

parlist

list of parameters to pass to distribution functions

ranges

lower, middle, and upper values to bracket lower and upper boundaries of the credible interval

level

confidence level

eps

if ranges is missing, set lower and upper brackets to the eps and 1-eps quantiles of the distribution

tol

tolerance on credible interval

verbose

if TRUE, return detailed information on the probability cutoff and realized area of the credible interval; if FALSE, just lower and upper bounds of the credible region

pvec

numeric vector of parameter values

npost

numeric vector of posterior density values corresponding to pvec

Details

tcredint gives credible intervals for a theoretical posterior density with defined density, cumulative density, and quantile functions; ncredint gives credible intervals for a numerical posterior density.

Value

A numeric vector giving the credible interval. If verbose=FALSE, gives just lower and upper bounds; if verbose=TRUE, also gives information on the probability cutoff and realized area of the credible interval

Note

For credible intervals from a sample (e.g. from an MCMC run), see HPDinterval in the coda package.

Author(s)

Ben Bolker

Examples

tcredint("beta",list(shape1=5,shape2=10),verbose=TRUE)
pvec = seq(0,1,length=100)
postvec = dbeta(pvec,shape1=5,shape2=10)
ncredint(pvec,postvec,verbose=TRUE)
set.seed(1001)