Mixture of Experts (MR-MoE)

MR-MoE uses machine learning to choose the best performing model out of multiple pleiotropy-robust estimation methods. 

Based on meta-learning, in which simulations are used to train a machine which selects the method most likely to return a reliable MR estimate. 

References

• Hemani G, Bowden J, Haycock PC, et al. Automating Mendelian randomization through machine learning to construct a putative causal map of the human phenome. bioRxiv 2017: 173682.

Other terms in 'Model selection and averaging approaches for two-sample MR':

• Bayesian model averaging for IVW and MR-Egger
• Rucker framework
• Steiger filtering