znormix: Normal-mixture based estimation of LFDR and pi0

In pi0: Estimating the proportion of true null hypotheses for FDR

Description

This function implements the method of McLachLan, Bean and Jones (2006).

Usage

znormix(p, theoretical.null=TRUE, start.pi0, eps=1e-5, niter=Inf, verbose=FALSE)

Arguments

p	a numeric vector the p-values
theoretical.null	logical scalar, indicating whether theoretical N(0,1) null distribution is assumed for z-scores.
start.pi0	optional numeric scalar, starting value of pi0 for EM algorithm; if missing, qvalue will be called with default arguments to get this starting value.
eps	numeric scalar, maximum tolerable absolute difference of parameter estimates for successive iterations in the EM algorithm.
niter	numeric scalar, maximum number of EM iterations.
verbose	logical scalar, indicating whether excessive outputs will be printed during EM algorithm.

Details

A two-component normal mixture model is fit thru EM algorithm on the z-scores, where z=qnorm(1-p).

Value

A length 5 numeric named vector of estimated parameters, with class 'znormix' and attributes

theoretical.null	the same as input.
converged	logical, convergence status.
iter	numeric, number of iterations.
call	the match.call() result.
lfdr	numeric vector of local false discovery rates, with order being the same as the input p-values.
fdr	numeric vector of false discovery rates, with order being the same as the input p-values.

Note

There are two small differences with McLachlan, Bean and Jones (2006):

• If start.pi0 is missing, it is estimated by the q-value smoother method implimented in qvalue.

• For the empirical null case, a call to quantile(z, start.pi0) is used as the threshold to determine the initial component assignment, when choosing starting values.

Author(s)

Long Qu

References

G.J. McLachlan, R.W. Bean and L. Ben-Tovim Jones. (2006) A Simple implementation of a normal mixture approach to differential gene expression in multiclass microarrays. Bioinformatics, 22(13):1608-1615.

See Also

qvalue, histf1

Examples

set.seed(9992722)
p=1-pnorm(c(rnorm(7000),rnorm(3000,1)))
znormix(p, eps=1e-5)['pi0'] ##  0.6759555

pi0 documentation built on May 2, 2019, 4:47 p.m.