Genetic Correlation - Confidence Intervals

[Nils Kappelmann]

Dear LDSC team,

I was wondering how you would compute confidence intervals from LDSC genetic correlation results:

Would you, for instance, simply use the following for the common z cutoff for 95% CI:

CI = rg ± 1.96*se

Or would you alternatively use Fisher's z transformation as is commonly done for correlations:

zprime = 0.5 * log((1+rg) / (1-rg))

ci = zprime ± 1.96*se

ci  = ((exp(2*ci) - 1) / (exp(2*ci) + 1))

If using Fisher's z transformation, I'm also not completely sure how to transform the standard error and/ or whether this has already been transformed.

I couldn't find a similar question in the google group but my apologies if I missed it somewhere and this is redundant.

Thanks in advance for the support and response!

Best wishes,

Nils