Question about burden inflation versus skat-o

[oalavijeh]

Dear SAIGE team,

Many thanks for this amazing tool! We ran SAIGE-GENE genome-wide on a binary phenotype with a strong known LoF mechanism but were open to the idea of finding variants of all effects (1200 cases vs 25000 controls, WGS). Our results shows that the burden test shows lower p-values than SKAT-O for all positive controls across masks, but also inflation. I was wondering why this could potentially be? Our variants are all of high quality, we controlled for cryptic relatedness even before input with KING and batch effects. Does the burden test use slightly different genomic controls or do I need to do some more digging in out data?

All the best