Open jjandshi opened 1 year ago
Hi @jjandshi,
to make sure I understand this issue, in both analyses, the exposure and outcome are the same, correct?
There are several reasons that could explain your results. 1) Indeed, MR-PRESSO is based on statistical tests, and as you increase your number of IVs, corresponding to observations in a classical test and linked to the residual degrees of freedom, you increase your statistical power. 2) MR-PRESSO detects outliers by comparing a specific SNP to all the other SNPs in your analysis. Therefore, it might be possible that by including these 10 extra SNPs, the outlier appears even more as an outlier and is detected in this context.
Best, Marie
I have selected two sets of instrumental SNPs from index SNPs of a large GWAS for MR-PRESSO, with numbers of 30 and 40 respectively. Both sets include one same SNP with strongest effect on exposure (beta= 0.26) and weak effect on outcome (beta=0.04, and p=0.002). However, MR-PRESSO did not detect this SNP as a outlier for the first dataset( RSSobs and Pvalue not shown?), but for the second dataset (RSSobs=0.0023, Pvalue=0.004). Does any one have idea on this issue? Thanks, Justin