Quantile normalization

[Katherine Fawcett]

Hello,

When I use the quantile-normalization option in Plink I get different results to if I simply input the trait already quantile normalized in R using the following method:

> df$pheno.rank <- rank(df$pheno)

> df$pheno.transformed <- qnorm((df$pheno.rank-0.5)/length(df$pheno.rank))

I was wondering exactly what the quantile-normalization parameter does to phenotypes/covariates that is different to this.

Thanks

Kath

[Christopher Chang]

Can you provide an example of something plink2 appears to be doing wrong?  I just tested this and the results were identical between R and plink2.