allow alternative thresholds for outlier detection

yoonsucho commented 4 years ago

Alternative thresholds have been implemented.

[ ] In some cases, there are different numbers of outliers identified when I use tryx.scan and RadialMR respectively. For example, I have ran tryx.scan and have manually perform RadialMR to check if the commands for RadialMR work fine in MR-TRYX. Generally the commands work fine. However, in the analysis of urate-eGFR, tryx.scan identified 10 outliers when Bonferroni correction was applied whereas RadialMR detected 9 outliers. I found that Q statistics and P values were estimated differently depending on whether I use RadialMR directly or use RadialMR within MR-TRYX.

explodecomputer commented 4 years ago

The Q statistics seem to be slightly different - is the same number of SNPs being used in both examples? In tryx i think it might remove the ones that have mr.keep=FALSE or something like that?

yoonsucho commented 4 years ago

Sorry, I think there was a minor error in my commands. yes you are right. Thanks for pointing this out:-)

explodecomputer / tryx

allow alternative thresholds for outlier detection #6