Seems like p-values for the pm has changed in the past half a year

kaz-yos / regmedint

R implementation of effect measure modification-extended regression-based closed-formula causal mediation analysis

https://kaz-yos.github.io/regmedint/

28 stars 5 forks source link

Seems like p-values for the pm has changed in the past half a year #10

Open dianashams opened 2 years ago

dianashams commented 2 years ago

Hello, Thank you for the package, it seemed to be one of the very few (only?) options for survival outcomes in R. I wonder if something happened between Dec21/jan22 and now Aug22? I am re-running analyses from that time and all the results from redmedint function are the same but the confidence interval for partial mediation (much wider now). Both for logistic mreg and yreg, and for survCox yreg.

einsley1993 commented 1 year ago

Hi @dianashams , sorry for the late response. There was indeed some little changes in the variance matrices that resulted in change in 95% CI and P-value between 2021 and 2022. Earlier we had a typo in variance functions and have now fixed it.