FESAT: Fixed-effects (FE) meta-analysis of variant-set association

In baolinwu/MetaSAT: Meta-analysis of variant-set association tests

Description

Conduct meta-analysis of variant-set association test of m variants assuming constant effects across K studies. These association statistics are typically Score vector, and a direct summation asymptotically amounts to inverse variance weighting. In practice, we typically input weighted test statistics. FE VT: (∑_kU_k)^T(∑_kU_k); FE BT: (∑_kη^TU_k)^2; FAT: adaptively combine FE VT and FE BT.

Usage

FESAT(Us, Vs, eta = NULL, rho = (0:10/10)^2)

Arguments

Us	matrix of variant test statistics (m by K)
Vs	array of covariance matrix for test statistics (mxm by K)
eta	coefficient vector for the FE BT. Default to equal weights.
rho	weights assigned to the FE BT

Value

p.value

p-values for FAT, FE VT, FE BT

pval

vector of p-values for each rho

rho.est

optimal rho value leading to the minimum p-value

References

Wu,B. and Zhao,H. (2018) Efficient and powerful meta-analysis of variant-set association tests using MetaSAT.

Examples

K = 3; m=10
Vs = array(0, dim=c(m,m,K)); Us = matrix(0, m,K)
for(k in 1:K){
  ak = matrix(rnorm(100*m),100,m)*sqrt(0.8)+rnorm(100)*sqrt(0.2)
  Vs[,,k] = cor(ak)
  Rh = chol(Vs[,,k])
  Us[,k] = colSums(Rh*rnorm(m))
}
FESAT(Us,Vs)
U1 = Us + runif(m*K, 0,2)
FESAT(U1,Vs)

baolinwu/MetaSAT documentation built on May 13, 2019, 5:22 p.m.