lik_pois: Likelihood object for Poisson error distribution

In ashr: Methods for Adaptive Shrinkage, using Empirical Bayes

Likelihood object for Poisson error distribution

Description

Creates a likelihood object for ash for use with Poisson error distribution

Usage

lik_pois(y, scale = 1, link = c("identity", "log"))

Arguments

y	Poisson observations.
scale	Scale factor for Poisson observations: y~Pois(scale*lambda).
link	Link function. The "identity" link directly puts unimodal prior on Poisson intensities lambda, and "log" link puts unimodal prior on log(lambda).

Details

Suppose we have Poisson observations y where y_i\sim Poisson(c_i\lambda_i). We either put an unimodal prior g on the (scaled) intensities \lambda_i\sim g (by specifying link="identity") or on the log intensities log(\lambda_i)\sim g (by specifying link="log"). Either way, ASH with this Poisson likelihood function will compute the posterior mean of the intensities \lambda_i.

Examples

   beta = c(rnorm(100,50,5)) # prior mode: 50
   y = rpois(100,beta) # simulate Poisson observations
   ash(rep(0,length(y)),1,lik=lik_pois(y))

ashr documentation built on Aug. 22, 2023, 1:07 a.m.