corMeta.GLS.1: The efficient meta-analysis estimator accounting for...

In tamartsi/MetaCor: Performs Stratified SNP-trait analysis and combines the results using the MetaCor method

Description

Computes the fixed effects meta-analysis estimator while accounting for correlations between effect estimates in the variance estimation. The computation is done for a single SNP at a time.

Usage

corMeta.GLS.1(beta.1, var.1, cov.1)

Arguments

beta.1	A vector of S effect estimates
var.1	A vector of S variances of the effect estimates in beta.1
cov.1	A vector of S*(S-1)/2 covariances between the effect estimates in beta.1

Details

Computes the MetaGLS (MetaCor1 form the MetaCor paper) estimator and test for a single SNP.

Value

A data frame with 1 row, with beta, var, test.stat and pval providing the meta-analysis effect estimate, its variance, test statistic, and p-value.

Note

This function is called by MetaCor and it is advisable to use MetaCor. It can also be called directly, but then the test of heterogeneity is not performed. Also, MetaCor allows for selection of strata while this function does not.

Author(s)

Tamar Sofer

Examples

require(mvtnorm)
n.strat <- 5
n.est <- 5
## create a positive definite matrix to be a covariance matrix
sigma <- matrix(rnorm(n.strat^2), n.strat, n.strat)
sigma <- t(sigma) 

effects <- rmvnorm(n.est, c(2,2,3,3,0), sigma)
colnames(effects) <- paste0("Beta.s", 1:n.strat)
vars <- matrix(rep(diag(sigma), n.est), nrow = n.est, ncol = n.strat, byrow = T)
colnames(vars) <- paste0("var.s", 1:n.strat)
covs <- matrix(rep(sigma[lower.tri(sigma)], n.est), nrow = n.est, byrow = T)
colnames(covs) <- c(paste0("cov.s1:", paste0("s", 2:n.strat)), paste0("cov.s2:", paste0("s", 3:n.strat)), paste0("cov.s3:", paste0("s", 4:n.strat)) , "cov.s4:s5")


corMeta.GLS.1(effects[1,], vars[1,], covs[1,], testType = "MetaGLS")

tamartsi/MetaCor documentation built on May 31, 2019, 2:56 a.m.