Is the distribution of effect size estimates specifically and strictly required to follow Normal Distribution?
EthanZhaoChem opened this issue · 0 comments
Hi @chr1swallace ,
Many thanks for building such a useful package with various input considerations and up-to-date coordination with SuSiE.
My question is whether the effect size estimates, which could be used as input for coloc, must follow normal distribution. I noticed you mentioned in this issue that beta is assumed to come from a normal distribution with mean 0 and standard deviation 0.15 * sdY. However, in our analysis, QTLs were called using RASQUAL, which doesn't assume effect size to be normal distribution. So I am wondering whether these effect estimates are still proper to be used for coloc as inputs.
Sometimes even if under normal distribution assumption, I think it's also probable that people still get effect estimates that are not strictly following normal distribution. So if the assumption (of distribution) or distribution itself is strictly required to be normal, will a pre-normalization step be helpful?
Best,
Ethan