MR Dictionary

Contamination mixture models

Method that uses a contamination mixture model, which aims to construct and maximise a likelihood function based on the instrumental variable (IV)-specific causal effect estimates in a two-sample MR setting.

If a genetic IV is valid (i.e., does not violate any MR assumptions), then its causal effect estimate will be normally distributed about the true value of the causal effect. The method can robustly and efficiently estimate the causal effect of an exposure on an outcome, even if some of the genetic IVs are invalid, or can identify distinct groups of genetic variants that have similar causal effect estimates. In the latter scenario, this may imply several biological mechanisms by which the exposure influences the outcome.

References

Other terms in 'Pleiotropy-robust two-sample MR methods':