I want to know if I am trying to do a 2 stage mendelian randomization. Using R software with the mendeliazation package,
Since using a logistic regression the snp does not have a significant p value with the outcome Mi but with a Cox proportional model it does. How then can I find a causal effect or effect size of the exposure on the outcome.
My lecture said I could divide the coef of the log hazard of the snp to the outcome from the Cox by the coefficient of the linear regression of the snp and the exposure to get my effect size. I still didn't understand how I could that.
Also when I went ahead to use the predict model to get unconfounding values for hdl and ldl, I realised that when regressed the uncounfounded model on the other outcome, I got 6 but when I did the coef of the log ( snp_ outcome effect) divided by the coefficient of the linear model (snp_ exposure effect) I got 0.069.
I have been finding it difficult to reconcile these two.
Will be glad if someone can help me answers to this.
Thanks
Check out this article and let me know if you still need help.
https://www.nature.com/articles/s41467-019-14156-4
- Badges
- Science topic
- Similar topics
- Comment