2 p.m. - 3 p.m. Location: JO 4.614
Michael B. Sohn
Department of Biostatistics and Epidemiology
University of Pennsylvania
Statistical Methods in Microbiome Data Analysis
Microbiome study involves new computational and statistical challenges due to unique characteristics of microbiome data: highly sparse, skewed, over-dispersed, and high-dimensional. In this talk, I will present two methods: 1) a GLM-based latent variable ordination method to resolve potential problems of the distanced-based ordination method, such as principal component analysis (PCoA), under strong dispersion effect, and 2) a compositional mediation model to extend the applicability of mediation analysis to a model with high-dimensional compositional mediators. The proposed mediation model utilizes the algebraic structure of composition under the simplex space and a constrained linear regression model to achieve subcompositional coherence. The methods will be illustrated with Penn Upper Respiratory Tract Microbiome Dataset and Penn Gut Microbiome Project - COMBO Dataset.
Sponsored by the Department of Mathematical Sciences