Biostatistics Journal Club: Directional Testing in Mediation Analysis
This discussion, led by Jonathan Dreyfuss, PhD, Joslin Diabetes Center, will introduce the causal mediation setting and current approaches to it. Using a motivating example in bioinformatics, Dreyfuss will discuss his approach (termed “Hitman”) that accounts for the direction of effects of the mediators. He will show via simulations that it provides stronger p-values than current approaches and show the results on the motivating example, which were confirmed experimentally. Work since the publication on providing stronger adjusted p-values (either FDR or FWER) will also be discussed.