question about the outliers

[adamber6]

Dear authors,
I appreciate your precious time.
I want to ask a question. I have performed the MR-PRESSO analysis about 40 times (the SNP is 1000, and the NbDistribution = 2000), and I deleted the possible outliers. However, the global test is still less than 0.05. Can you tell me how to solve the problem. My code is as following:

mr_presso(BetaOutcome ="beta.outcome",

•       BetaExposure = "beta.exposure", 
  

•       SdOutcome ="se.outcome", 
  

•       SdExposure = "se.exposure", 
  

•       OUTLIERtest = TRUE,
  

•       DISTORTIONtest = TRUE, 
  

•       data =Mydata1, 
  

•       NbDistribution = 2000, 
  

•       SignifThreshold = 0.05)
  

[marieverbanck]

Hi @adamber6,
Could you clarify a couple of elements to make sure we understand your issue.
Are you saying that you conducted 40 MR-PRESSO analyses with the same data and iteratively removed outliers?
If so, you should be extra careful on how to control your false positive rate in that situation, because of the multiplicity of the tests you are doing.
On a different note, you mentioned 1000 SNPs which could explain why you have power to detect even smaller pleiotropic effects. BTW, 1000 SNPs seem quite high, are they pruned/clumped? How many exposures do you have in your model?
Best,
Marie