Should latent variable variance be set to 1 for downstream analyses?

[Thanh Le]

Hello,

I’m currently performing post-GWAS analyses using multivariate GWAS summary statistics generated from GenomicSEM.

For post/GWAS analyses that require both summary statistics and sample size, I have been using the summary statistics from GenomicSEM (with std.lv = TRUE) together with the implied sample size calculated from the summary statistics run with std.lv set to FALSE.

I was wondering whether it would be appropriate to use the summary statistics generated from userGWAS() with std.lv = TRUE together with this implied sample size for downstream analyses, for example, stratified LDSC.

Thank you very much for developing and maintaining such an excellent tool. I look forward to your insights.

Best regards,
Thanh Le

[Elliot Tucker-Drob]

That depends on exactly what you are using those betas and Ns for. If you set the factor SD to 1, then the betas and their SEs will be scaled relative to what is essentially a 100% heritable phenotype, which means that the implied N will appear to be very small (for highly heritable phenotypes, power is very high, i.e. a smaller N is needed to obtain the same level of precision). You will generally find more intuitive Ns when using unit loading identification rather than unit variance identification. Unit loading will put the implied N on essentially the same scale as the reference indicator for which the unstandardized loading has been fixed to 1.0. Both identification strategies, if correctly implemented, should produce the same standardized results, and the Z statistics for the betas (and thus the p values) should be extremely similar.

 

--
You received this message because you are subscribed to the Google Groups "Genomic SEM Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to genomic-sem-us...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/genomic-sem-users/d515e5a2-6a48-4d95-b9eb-1b79246a5780n%40googlegroups.com.