compute_sample_size: Compute (effective) sample size

In neurogenomics/MungeSumstats: Standardise summary statistics from GWAS

Compute (effective) sample size

Description

Computes sample sum (as new column "N") or effective sample size (ESS) (as new column "Neff"). Computing ESS is important as it takes into account the proportion of cases to controls (i.e. class imbalance) so as not to overestimate your statistical power.

Usage

compute_sample_size(
  sumstats_dt,
  method = c("ldsc", "giant", "metal", "sum"),
  force_new = FALSE,
  append_method_name = FALSE
)

Arguments

sumstats_dt	Summary statistics data.table.
method	Method for computing (effective) sample size. "ldsc" : Neff = (N_CAS+N_CON) * (N_CAS/(N_CAS+N_CON)) / mean((N_CAS/(N_CAS+N_CON))[(N_CAS+N_CON)==max(N_CAS+N_CON)])) bulik/ldsc GitHub Issue bulik/ldsc GitHub code "giant" : Neff = 2 / (1/N_CAS + 1/N_CON) Winkler et al. 2014, Nature "metal" : Neff = 4 / (1/N_CAS + 1/N_CON) Willer et al. 2010, Bioinformatics "sum" : N = N_CAS + N_CON Simple summation of cases and controls that does not account for class imbalance. "\<integer\>" : N = \<integer\> If method is a positive integer, it will be used as N for every row.
force_new	If "Neff" (or "N") already exists in sumstats_dt, replace it with the recomputed version.
append_method_name	should Neff column have an indicator to explain the method that makes it., Default is FALSE unless multiple methods are passed

Details

There are many different formulas for calculating ESS, but LDSC is probably the best method available here, as it doesn't assume that the proportion of controls:cases is 2:1 (as in GIANT) or 4:1 (as in METAL).

Value

A data.table with a new column "Neff" or "N"

neurogenomics/MungeSumstats documentation built on Aug. 10, 2024, 5:59 a.m.