FDRforTailPP: FDR for tail posterior probability

In BGmix: Bayesian models for differential gene expression

Description

Calculate the false discovery rate (FDR) for the tail posterior probability

Usage

FDRforTailPP(tpp, a1, a2 = NULL, n.rep1, n.rep2 = NULL, prec = 0.05, p.cut = 0.7, N = 10000, pp0=NULL, plot = T)

Arguments

tpp	vector of tail posterior probabilities
a1	posterior mean of the shape parameter of the inverse gamma distribution - prior for the variance in condition 1
a2	posterior mean of the shape parameter of the inverse gamma distribution - prior for the variance in condition 2
n.rep1	number of replicates in condition 1
n.rep2	number of replicates in condition 2
prec	precision of the estimate of the cumulative distribution function of tail posterior probability under H0 (at points 1 - k*prec, k =1,2,..)
p.cut	to save time, calculate FDR only for cutoffs on tail posterior probability > p.cut
N	simulation size for tail posterior probability under H0
pp0	a vector of simulated tail posterior probabilities under H0
plot	if True, the estimated pi0 at different locations and the median estimate is plotted

Value

pi0	estimate of pi0 - proportion of non-differentially expressed genes
FDR	estimate of FDR for all (distinct) cutoffs > p.cut

Author(s)

Natalia Bochkina

References

Bochkina N., Richardson S. (2007) Tail posterior probability for inference in pairwise and multiclass gene expression data. Biometrics.

See Also

TailPP, FDRplotTailPP,histTailPP,EstimatePi0

Examples

 data(ybar, ss)
 nreps <- c(8,8)

## Note this is a very short MCMC run!
## For good analysis need proper burn-in period.
 outdir <- BGmix(ybar, ss, nreps, jstar=-1, nburn=0, niter=100, nthin=1)

 params <- ccParams(outdir)  
 res <-  ccTrace(outdir)
  
 tpp.res <- TailPP(res, nreps, params, plots  = FALSE)
 FDR.res = FDRforTailPP(tpp.res$tpp, a1 = params$maa[1],
a2 = params$maa[2], n.rep1=nreps[1], n.rep2=nreps[2], p.cut = 0.8)

BGmix documentation built on Nov. 8, 2020, 4:54 p.m.