MR Dictionary

Leave-one-out analysis

A method that performed MR analyses with each single nucleotide polymorphism (SNP) or instrumental variable (IV) is iteratively removed.

Traditionally used as a sensitivity analysis, this method enables the assessment of whether or not any causal effect estimate is being driven by one SNP independently of all other SNPs being used as IVs by comparing the original causal effect estimate (i.e., with all SNPs included in the model) to estimates where each SNP has been removed. If the causal effect estimates attenuates towards the null upon the removal of one SNP, this suggests that the original causal effect estimate may be biased by that individual SNP (i.e., through horizontal pleiotropy). Developed for two-sample MR settings but can be applied within one-sample MR settings.

References

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