clfdr: Conditional Local False Discovery Rate
In pkimes/otherfdr: Other FDR Methods

Description Usage Arguments Details Value Author(s) References Examples

An implementaiton of the conditional local false discovery rate procedure of Cai and Sun (2009). This is a near copy of the implementation provided in the IHWpaper package, clfdr at the corresponding GitHub repo, written by Nikos Ignatiadis and made available under the GPL-3 license.

The original code has been modified to only return a vector of adjusted p-values and check the input of the lfdr_estimation= parameter.

1	clfdr(p, groups, lfdr_estimation = c("fdrtool", "locfdr"))

`p`	numeric vector of unadjusted p-values.
`groups`	factor to which different hypotheses belong.
`lfdr_estimation`	method used to estimate the local fdr; must be one of "fdrtool" or "locfdr". (default = "fdrtool")

If the covariate or variable for stratifying p-values is continuous, it should be converted into a discrete set of "groups" before applying this function. This can be accomplished by passing the continuous variable to IHW::groups_by_filter, ggplot2::cut_number, or any similar function. See examples for uses of these functions.

Numeric vector of adjusted p-values of equal length and order as the to input vector of p-values.

Patrick Kimes

Reference for the conditional local false discovery rate (underlying theory):

Cai, T.T., Sun, W. (2009) Simultaneous testing of grouped hypotheses: Finding needles in multiple haystacks. Journal of the American Statistical Association, 104(488):1467-1481. https://doi.org/10.1198/jasa.2009.tm08415

Reference for the original implementation (underlying code):

Ignatiadis, N. (2017) IHWpaper: Reproduce figures in IHW paper. R package version 1.7.0. https://doi.org/doi:10.18129/B9.bioc.IHWpaper

## generate example p-values and covariate values
pv <- runif(100)
x <- sample(100)

## discretize 'x' into 10 bins
xd <- IHW::groups_by_filter(xd, nbins = 10)

## adjust p-values, grouping by covariate bin
adjpv <- clfdr(p = pv, groups = xd)