BGmix (version 1.32.0)

calcFDR: Estimate the FDR (false discovery rate) and related quantities for BGmix output.

Description

Given a threshold on the posterior probabilities, genes are declared as null or differentially expressed. For any given threshold, the FDR (false discovery rate) and FNR (false non-discovery rate) can be estimated using the posterior probabilities. Estimated numbers of false positives and false negatives are also output.

Usage

calcFDR(res, pcut = seq(0.01,0.5,0.01), true.z = NULL, q.print = F)

Arguments

res

list object output from ccParams (this includes the posterior classification probabilities)

pcut

scalar or vector of thresholds for which to estimate FDR etc.

true.z

vector of true classifications (if known, eg. for simulated data)

q.print

Print FDR etc. when pcut is a vector?

Value

fdr.est, fnr.est

scalars or vectors of estimated FDR, FNR

fp.est, fn.est

scalars or vectors of estimated no. false positives, no. false negatives

fdr.true, fnr.true

scalars or vectors of true FDR, FNR

fp.true, fn.true

scalars or vectors of true no. false positives, no. false negatives

npos, nneg

scalars or vectors of no. declared positives, no. declared negatives

prob.class

posterior classification probabilites (from the 'res' object input to this function)

true.z

argument to function is output

pcut

argument to function is output

Details

If the true classification is known, it can be given as true.z, and the true FDR etc. for the threshold probability can be calculated.

Examples

## Note this is a very short MCMC run!
## For good analysis need proper burn-in period.
data(ybar,ss)
outdir <- BGmix(ybar, ss, c(8,8), nburn=0, niter=100, nthin=1)
params <- ccParams(outdir)
fdr <- calcFDR(params)