prop.alt {unknown}R Documentation

Internal function for FDR multiple testing

Description

This is an internal function that calculates an estimate of a, the proportion of alternative hypotheses, using one of several methods.

Usage

prop.alt(pvals,adjustment.method="mean",adjustment.args=list(edf.lower=0.8,num.steps=20))

Arguments

pvals a vector of pvals on which to conduct the multiple testing.
adjustment.method an optional character string indicating the method for increasing the power of the procedure by estimating the proportion of alternative p-values, one of "mean", the modified Storey estimator that we suggest in Ventura et al. (2004), "storey", the method of Storey (2002), or "two-stage", the iterative approach of Benjamini et al. (2001)
adjustment.args a list giving the arguments to adjustment.method:
for adjustment.method="mean", specify edf.lower, the smallest quantile at which to estimate a, and num.steps, the number of quantiles to use - the approach uses the average of the Storey (2002) estimator for the num.steps quantiles starting at edf.lower and finishing just less than 1
for adjustment.method="storey", specify edf.quantile, the quantile at which to calculate the estimator
for adjustment.method="two-stage", the method uses a standard FDR approach to estimate which p-values are significant; this number is the estimate of a; therefore the method requires specification of qlevel, the proportion of false positives and fdr.method ('original' or 'general'), the FDR method to be used. We do not recommend 'general' as this is very conservative and will underestimate a.

Value

estimate of a, the number of alternative hypotheses

Author(s)

Chris Paciorek paciorek AT alumni.cmu.edu

References

see fdr.

Examples

a <- prop.alt(pvals,adjustment.method="mean",adjustment.args=list(edf.lower=0.8,num.steps=20))

a <- prop.alt(pvals,adjustment.method="storey",adjustment.args=list(edf.quantile=0.8))

a <- prop.alt(pvals,adjustment.method="two-stage",adjustment.args=list(qlevel=0.05,fdr.method='original'))

[Package Contents]