MR Dictionary

Statistical power and efficiency

In many situations, MR will be less biased than conventional multivariable regression (or similar) approaches to exploring causal relationships with observational data but will have less statistical power. 

Large consortia used for MR studies and the increasing availability of publicly available summary statistics from genome-wide association studies (GWASs) that can be used in two-sample MR can provide greater (and often adequate) statistical power for detecting hypothesized effect sizes. 

References

Other terms in 'Sources of bias and limitations in MR':