Omnibus Testing Using Summary Statistics from Biobanks

June 2020 – October 2020

Overview

This project, headed by Dr. Nathan Tintle, developed a mathematical framework for computationally efficient, exact multimarker omnibus tests using individual marker summary statistics from large biobanks.

Research Approach

Using matrix algebra, we demonstrated that various omnibus tests could be conducted using only summary statistics provided by biobanks.
Using simulations, we demonstrated the accuracy of the proposed methods and evaluated variance attributable to randomization.
Using omega-3 fatty acid data from several cohorts of the Framingham Heart Study, we validated the methods and demonstrated high accuracy with low variance.

Figures

Simulation Results Framingham Heart Study Analysis

Research Outputs

Abstract Published Paper