REma: Random effects (RE) meta-analysis (MA) of multiple...

In baolinwu/GSmeta: Various meta-analysis methods with main applications to genetic studies

Description

Under RE MA, we assume heterogeneous effect sizes across all studies. This is potentially relevant when meta-analyzing cross-disease GWAS with overlapped samples.

Usage

REma(betas, Vb, B = 2500)

Arguments

betas	effect size estimates from K studies
Vb	covariance matrix of effect size estimates
B	number of Monte Carlo samples to estimate the null distribution. Default to 2500. See Wu and Zhao (2018) reference.

Value

p.value

p-values for FE (Pf), RE (Pr), RE conditional on FE (Pc)

theta

proportion parameter in the chi-square mixture dist

Q

LRT statistics for the FE (Qf) and RE (Qr)

pars

estimated parameters (muf,mu,tau2) for FE mean, RE mean/variance parameters

References

Lin,D.Y. and Sullivan,P.F. (2009) Meta-Analysis of Genome-wide Association Studies with Overlapping Subjects. Am J Hum Genet 85, 862<e2><80><93>872.

Han,B. and Eskin,E. (2011) Random-Effects Model Aimed at Discovering Associations in Meta-Analysis of Genome-wide Association Studies. The American Journal of Human Genetics 88, 586<e2><80><93>598.

Lee,C.H., Eskin,E., Han,B. (2017) Increasing the power of meta-analysis of genome-wide association studies to detect heterogeneous effects. Bioinformatics 33, i379<e2><80><93>i388.

Wu,B. and Zhao,H. (2018) Powerful random effects modeling for meta-analysis of genome-wide association studies.

Examples

K = 5; rho = 0.2
V = diag(K)*(1-rho) + rho
U0 = rnorm(K)*sqrt(1-rho) + rnorm(1)*sqrt(rho)
U1 = U0 + rnorm(K)*1.2
REma(U0,V)
REma(U1,V)

baolinwu/GSmeta documentation built on May 24, 2019, 7:13 a.m.