Question About qval vs. pval_beta in TensorQTL Permutation Results

[MinS1]

Hi,

I am using TensorQTL to analyze conditional cis-QTLs, and I observed an unusual behavior in the permutation results. After applying post.calculate_qvalues to calculate q-values from the permutation results, I noticed that many qval values are smaller than their corresponding pval_beta.

From my understanding, qval should typically be greater than or equal to pval_beta due to the multiple testing correction process. Could you please confirm if this behavior is expected? If not, do you have any suggestions for troubleshooting or steps to verify whether the calculations were performed correctly?

Thank you for your assistance!

Best regards,
Shi

[francois-a]

Hi, how does the pval_beta histogram/QQ plot look like? This can happen when there is only a small proportion of null data, in which case the pi0 estimate may be inaccurate. It may be preferable to use Benjamini-Hochberg in those cases.