Post by Jorien on May 25, 2020 8:14:28 GMT
I have a question about the standardization step of GSMR as described at cnsgenomics.com/software/gsmr/
The instructions state that the input should be allele frequencies, z-statistics and sample size. Could you please clarify what is meant exactly with
z-statistics here? I assume it is not standardized, z-scores, as the script also requires you to include SEs.
In my case, the unit of the effect size I have is an unstandardized beta (a volume measure), would it be correct if I input these unstandardized betas and use
the standardization step to get to a standardized beta?
Thank you in advance!
Post by ax on Jun 22, 2020 6:00:53 GMT
The standardizaiton step is only for beta and se estimate, while Z-statistic is already a standardized statistic.
When you standardize the estimate, the actual process is
1. You input allele frequencies, z-statistics and sample size
2. Calculate the standardized se by se_new = 1 / sqrt(2*snp_freq*(1-snp_freq)*(n+zscore^2))
3. Calculate the standardized beta by beta_new= z / se_new
So you could see that z didn't change after the standardization process.
Hope this answer your question