Title: 'Simulated Sample Splitting (MR-SIMSS): A method to eliminate the effect of winner’s curse in Mendelian Randomisation’
Abstract: Mendelian randomization (MR) is a popular statistical technique that uses genetic variants to explore causal relationships in observational epidemiology. Summary-level MR, the most common form, relies on published GWAS summary statistics to estimate causal effects between exposures and outcomes. However, such analyses often ignore issues relating to Winner’s Curse of instrument effects, weak instrument bias and sample overlap, mechanisms that are known to induce substantial bias. In this talk, I describe a new method, MR Simulated Sample Splitting (MR-SimSS), designed to remove the effect of Winner’s Curse from summary-level MR. It operates by simulating statistically independent sets of summary statistics that are analogous to what would be produced by splitting individual-level data into independent subsets. These simulated summary statistics can then be plugged into existing methods that are robust to weak instrument bias and certain types of pleiotropy. In its 3-split form, MR-SIMSS will also remove bias due to overlap between the exposure and outcome GWAS datasets (even if they overlap completely), allowing MR to be applied in this setting with preserved Type I error.
Biography: John Ferguson is a Senior Lecturer in Statistical Science at the University of Galway. He holds a Ph.D in Statistics from Yale University (2009) and was previously a recipient of a Health Research Board Ireland Emerging Investigator Award. His research interests include Bayesian statistics, medical statistics, statistical genetics and causal inference.
________________________________________________________________________________
Microsoft Teams meeting
Join: https://teams.microsoft.com/meet/388850205917076?p=Q3XWM8UFoYZhAC3iVG
Meeting ID: 388 850 205 917 076
Passcode: PB2ns6VQ