Testing multiple biological mediators simultaneously.

ABSTRACT:

Motivation

Modern biomedical and epidemiological studies often measure hundreds or thousands of biomarkers, such as gene expression or metabolite levels. Although there is an extensive statistical literature on adjusting for 'multiple comparisons' when testing whether these biomarkers are directly associated with a disease, testing whether they are biological mediators between a known risk factor and a disease requires a more complex null hypothesis, thus offering additional methodological challenges.

Results

We propose a permutation approach that tests multiple putative mediators and controls the family wise error rate. We demonstrate that, unlike when testing direct associations, replacing the Bonferroni correction with a permutation approach that focuses on the maximum of

SUBMITTER: Boca SM 

PROVIDER: S-EPMC3892685 | biostudies-literature | 2014 Jan

REPOSITORIES: biostudies-literature