CpssFdrInfo: Estimate FDR thresholds based on Complementary Pairs...

In pjnewcombe/Pmisc: A collection of miscellaneous functions.

Description

Implements the FDR calculation developed by Shah and Samworth for FDR inferred by Complementary Pairs Stability Selection

Usage

CpssFdrInfo(cpss.selections, predictors = NULL, target.fdrs = NULL,
  selection.thresholds = NULL, q = NULL, fdr = TRUE)

Arguments

cpss.selections	A matrix of logical/binary indicators, or 0-1 probabilties if averaged over multiple imputation chains, of selection of variables across the complementary pairs subset analyses. Rows are variables, columns are subset analyses
predictors	Which subset of predictors to perform the FDR calculation over. Default NULL means all will be considered.
target.fdrs	A vector of FDR levels to calculate posterior probability thresholds for. Default NULL
selection.thresholds	Alternatively, a vector of selection probability thresholds, for which to calculate the FDR for, can be provided. Default NULL
q	Optional estimate of q - the true number of "signal" variables. If left NULL the sum of CPSS selection probabilities is used as a plug-in estimate (assuming the lambda's are cross-validated)
fdr	whether or not to return FDRs rather than "p-value" type measure

Value

fdr.info A list containing the target FDR rates, info for each including the posterior probability threhsold and the estimated FDR rate at that threshold across the permutations. The list of covariates reaching each threshold is also included

Author(s)

Paul Newcombe

pjnewcombe/Pmisc documentation built on March 26, 2020, 2:09 p.m.