Inflation observed for O.minp

[agilly]

Hi Han,

just a question w.r.t. the interpretation of results. I am running many parallel RVAS, and measuring the genomic inflation factor lambda as a quality check. While E.pval and O.pval are relatively well-controlled, I observe a very strong inflation for O.minp. When I calculate lambdas for p-values produced by MONSTER (also a GMM with O-test and rho-optimisation) I get no inflation.

Could you tell me which p-value I should use for the O-test? The documentation says that O.minp is the minimum p-value observed along the rho search grid. Is O.pval that p-value but corrected for the number of rhos tested?

Thanks,

Arthur

[hanchenphd]

Yes, O.minp is the minimum p-value observed from the search grid, and it does not follow a uniform distribution under the null hypothesis. O.pval is the p-value corrected for the search grid by numerical integration.

[agilly]

Great, thanks for the clarification!