GBJ: GBJ.R

In GBJ: Generalized Berk-Jones Test for Set-Based Inference in Genetic Association Studies

Description

Calculate the Generalized Berk-Jones test statistic and p-value.

Usage

GBJ(test_stats, cor_mat = NULL, pairwise_cors = NULL)

Arguments

test_stats	Vector of test statistics for each factor in the set (i.e. marginal test statistic for each SNP in a gene).
cor_mat	d*d matrix of the correlations between all the test statistics in the set, where d is the total number of test statistics in the set. You only need to specify EITHER cor_mat OR pairwise_cors.
pairwise_cors	A vector of all d(d-1)/2 pairwise correlations between the test statistics. You only need to specify EITHER cor_mat OR pairwise_cors.

Value

A list with the elements:

GBJ	The observed Generalized Higher Criticism test statistic.
GBJ_pvalue	The p-value of this observed value, given the size of the set and correlation structure.
err_code	Sometimes if your p-value is very small (<10^(-12) usually), R/C++ do not have enough precision in their standard routines to calculate the number accurately. In these cases (and very rarely others) we switch to standard Berk-Jones instead (more stable numerically) and let you know with a message here.

Examples

# Should return statistic = 0.9248399 and p_value = 0.2670707
set.seed(100)
Z_vec <- rnorm(5) + rep(1,5)
cor_Z <- matrix(data=0.2, nrow=5, ncol=5)
diag(cor_Z) <- 1
GBJ(test_stats=Z_vec, cor_mat=cor_Z)

GBJ documentation built on March 26, 2020, 6:05 p.m.