Learn R Programming

LocFDRPois (version 1.0.0)

Functions for Performing Local FDR Estimation when Null and Alternative are Poisson

Description

The main idea of the Local FDR algorithm is to estimate both proportion of null observations and the ratio of null and alternative densities. In the case that there are many null observations, this can be done reliably, through maximum likelihood or generalized linear models. This package implements this in the case that the null and alternative densities are Poisson.

Copy Link

Version

Install

install.packages('LocFDRPois')

Monthly Downloads

5

Version

1.0.0

License

GPL-2

Maintainer

Kris Sankaran

Last Published

May 5th, 2015

Functions in LocFDRPois (1.0.0)

BuildTable

Build a table corresponding to the histogram of data
AnalyticalOptim

Optimize lambda0 and pi0, using Poisson version of analytical approach in Efron's "Microarrays, Empirical Bayes, and the Two-Groups Model" In the paper, he derives formulas for arbitrary support. We use support [0, 1] for estimating the null. We initialize the search with lambda=0.05 and pi=0.95.
NullDensity

Estimate the density for the null data
LocfdrFuns

Run local fdr functions
LLConstructor

Return log-likelihood function giving Poisson version of equation (4.6) in "Microarrays, Empirical Bayes, and the Two-Groups model" by Efron.
SummarizeLocfdr

Call and summarize output to LocfdrFuns
MixtureDensity

Compute full mixture density for data, assuming exponential family model