process.f: process.f

In npetraco/fdrID: Computes local false discovery rate probabilities using JAGS, Stan and code from Efron, Turnbull and Narasimhan locfdr

Description

Computes estimates of the posterior null probabilities using f(x), the denominator of Bayes' Theorem in Efron's two-groups methodology.

Usage

process.f(zscores, xmids, pct0, lam.mat)

Arguments

zscores	Efron's z-scores derived from null and non-null score distributions
xmids	Midpoints of z-score histogram cells
pct0	Taken from locfdr package. Proportion of the z-score distribution used in fitting the null density f0(z) by central matching. If a 2-vector, e.g. pct0=c(0.25,0.60), the range [pct0[1], pct0[2]] is used. If a scalar, [pct0, 1-pct0] is used.
lam.mat	Point estimate(s) of f(x) (lambda(x) in the Poission regression). Each column of lam.mat is an extimate of f(x). In a Bayesian regression context, each column is a draw of f(x) from the joint posterior distribution.

Details

#Core function of the package. This algorithm take points making up the denominator of Bayes theorem in in Efron's methodology f(x), computes the likelihood function f0(x) = p(x|null) and (empirical) estimate of the prior p0. One or many estimates of f(x) can be contained (as columns) in lam.mat.

The algorithim is largely taken from locfdr code. Follows basically Efron 2006 Size, Power and False Discovery Rates

NOTE: Rows of lam.mat are f's generated by a Bayesian Poisson regression onto the hisogram of counts generated by z-scores.

Value

A list consisting of:

delta0, null mean estimates

sig0, null standard dev estimates

p0, null prior estimates

f(x) estimates.

References

locfdr package by Efron, Turnbull and Narasimhan. Archived at http://cran.r-project.org/src/contrib/Archive/locfdr/

Examples

XXXX

npetraco/fdrID documentation built on May 23, 2019, 9:33 p.m.