Instead of having pval3 as a column in output in ourInferenceData$sample_name, I got pval2.het.ind.. And after I went back and checked the difference between pval3 and pval2.het.ind., I found they are almost the same, except the case where the site is determined as heterozygous by joint genotyping but in that particular sample, the observed counts are highly biased (i.e. 0/10: ref/alt, pval2.het.ind. gives 1 but pval3 gives a significant result, and pval3 seems more reasonable to me).
If I understand it correctly, the likelihood test in QuASAR paper Eqn 9 needs genotype as prior. So, I am wondering if logLikBetaBinomialRhoEps needs to use the genotype computed by fitAseNullMulti.
If my guess is wrong, could you explain the meaning of pval2.het.ind.?
Instead of having
pval3as a column in output inourInferenceData$sample_name, I gotpval2.het.ind.. And after I went back and checked the difference betweenpval3andpval2.het.ind., I found they are almost the same, except the case where the site is determined as heterozygous by joint genotyping but in that particular sample, the observed counts are highly biased (i.e. 0/10: ref/alt,pval2.het.ind.gives 1 butpval3gives a significant result, andpval3seems more reasonable to me).If I understand it correctly, the likelihood test in QuASAR paper Eqn 9 needs genotype as prior. So, I am wondering if
logLikBetaBinomialRhoEpsneeds to use the genotype computed byfitAseNullMulti.If my guess is wrong, could you explain the meaning of
pval2.het.ind.?Thanks!