Decorrelated inverse-variance meta-analysis from summary statistics

When only summary statistics from independent target and source studies are available, the inverse-variance meta-analytic estimator \(\hat\beta^{meta}_j = w^{tar}_j\hat\beta^{tar}_j + w^{src}_j\hat\beta^{src}_j\) can serve as the auxiliary estimator. Because the two studies are independent, \(\hat\rho_j = w^{tar}_j\) and the decorrelated estimator is \(\hat\beta^{de}_j = w^{src}_j\hat\beta^{src}_j\) with variance \((w^{src}_j)^2 v^{src}_j\); its p-value therefore coincides with the source-study p-value. The output can be passed to heart_test() with p_de = p_de_meta.

Usage

heart_meta(beta_tar, se_tar, beta_src, se_src, snp = NULL)

Arguments

beta_tar, se_tar

Target-study estimates and standard errors.

beta_src, se_src

Source-study estimates and standard errors (aligned to the same effect allele).

snp

Optional variant identifiers.

Value

A data.frame with the target-only, meta-analytic and decorrelated estimates, standard errors and two-sided p-values.

Examples

heart_meta(beta_tar = c(0.1, 0), se_tar = c(0.05, 0.05),
           beta_src = c(0.08, 0.05), se_src = c(0.01, 0.01))
#>   snp beta_tar se_tar      p_tar  beta_meta     se_meta       p_meta   rho_meta
#> 1   1      0.1   0.05 0.04550026 0.08076923 0.009805807 1.767630e-16 0.03846154
#> 2   2      0.0   0.05 1.00000000 0.04807692 0.009805807 9.443044e-07 0.03846154
#>   beta_de_meta  se_de_meta    p_de_meta
#> 1   0.07692308 0.009615385 1.244192e-15
#> 2   0.04807692 0.009615385 5.733031e-07