lm_qvalue: Estimation of qvalues conditioned on covariates

In leekgroup/swfdr: Science-wise false discovery rate and proportion of true null hypotheses estimation

Description

The recipe for turning pvalues into qvalues is adapted from package 'qvalue' and articles by Storey, Tibshirani, Taylor, Siegmund.

Usage

lm_qvalue(
  p,
  X,
  pfdr = FALSE,
  pi0 = NULL,
  smoothing = c("unit.spline", "smooth.spline"),
  ...
)

Arguments

p	numeric vector of p-values
X	matrix of covariates (can be missing if pi0 is specified instead)
pfdr	logical, making estimates robust for small p-values and a small sample size
pi0	list with pi0 estimates from lm_pi0. If this is not provided, pi0 is estimated using function lm_pi0.
smoothing	character, type of smoothing used to fit pi0. Note the default in this function is different than in lm_pi0.
...	other parameters (passed on to lm_pi0 if pi0 is not provided)

Value

object of class ‘lm_qvalue’, which is a list with several components

call	matched function call
pvalues	numeric vector of original p-values
qvalues	numeric vector of q-values
	other list elements transferred from pi0

Examples

# define a covariate
X <- rep(c(0, 1), each=1000)
# generate p-values, randomly for group 0 and with low values for group 1
pVal <- c(runif(1000), rbeta(1000, 0.2, 1))
# compute qvalues, using the covariate
qVal <- lm_qvalue(pVal, X=X)

leekgroup/swfdr documentation built on Dec. 11, 2020, 11:40 a.m.