EBarrays (version 2.36.0)

crit.fun: Find posterior probability threshold to control FDR

Description

Find posterior probability threshold to control FDR

Usage

crit.fun(x, cc)

Arguments

x

x is one minus the posterior probabilities of being in a specific DE pattern. If there is only one DE pattern, then x is the posterior probabilities of being EE.

cc

cc is FDR to be controlled. For example, to control FDR at 0.05, set cc=0.05.

Value

• crit.fun returns a threshold so that if used in identifying genes in a specific DE pattern, FDR can be controlled at cc. Those genes with posterior probability of being in that specific DE pattern greater than this threshold are claimed to be in that specific DE pattern.

References

Newton, M.A., Noueiry, A., Sarkar, D., and Ahlquist, P. (2004). Detecting differential gene expression with a semiparametric hierarchical mixture model. Biostatistics 5, 155-176.

Examples

data(gould)
pattern <- ebPatterns(c("1,1,1,0,0,0,0,0,0,0",
                        "1,2,2,0,0,0,0,0,0,0"))
gg.em.out <- emfit(gould, family = "GG", hypotheses = pattern, num.iter = 10)
gg.post.out <- postprob(gg.em.out, gould)$pattern 
gg.crit <- crit.fun(gg.post.out[,1],0.05)
# number of DE genes 
sum(gg.post.out[,2] > gg.crit)

pattern4 <- ebPatterns(c("1, 1, 1, 1, 1, 1, 1, 1, 1, 1", 
                         "1, 2, 2, 2, 2, 2, 2, 2, 2, 2", 
                         "1,2,2,1,1,1,1,1,2,2", 
                         "1,1,1,1,1,1,1,1,2,2"))
gg4.em.out <- emfit(gould, family = "GG", pattern4, num.iter = 10)
gg4.post.out <- postprob(gg4.em.out, gould)$pattern
gg4.crit <- crit.fun(1-gg4.post.out[,2], 0.05)
# number of genes in pattern 2, a DE pattern
sum(gg4.post.out[,2] > gg4.crit)