What does it mean to calculate genomic control (lambda) at 10th percentile for GWAS

Question

[curious]

New to GWAS, been working on genomic control (GC,lambda). I see sometimes GC reported for the the 50th percentile, 10th percentile, 1st percentile, and 1/10th of a percentile.

I adapted code from this post: Calculating Genomic Inflation Factor

get_lambda <- function(in_path) {

    S <- fread(in_path)
    z <- qnorm(S$P.VALUE / 2)

    lambda <- round(median(z^2) / 0.454, 3)
    return(lambda)
}

This seems to replicate lambda for the 50th percentile, how would I go about getting the other percentiles?

Answer 1

[Kevin Blighe]

I also provided an associated answer here: Compute confidence intervals for lambda genomic factor statistics

The 'percentile' part would be related to the value of 0.454 in your code. This value of 0.454 represents the 50th percentile with 1 degree of freedom:

qchisq(0.5, 1)
[1] 0.4549364

For the 10th percentile, it would be:

qchisq(0.1, 1)
[1] 0.01579077

This being said, I would encourage you to run this by a 'card-carrying' statistician, and/or make a post on CrossValidated: https://stats.stackexchange.com/

Kevin