SDS Seminar Series – Matt Koslovsky, Colorado State University

The Fall 2024 SDS Seminar Series continues on October 25th from 2:00 p.m. to 3:00 p.m. with Dr. Matt Koslovsky (Assistant Professor, Department of Statistics, Colorado State University). This event is in-person in the Avaya Room (POB 2.302).    

Title: Sparse Dirichlet-Multinomial Models

Abstract: The Dirichlet-multinomial (DM) distribution and its variants are widely used to model multivariate count data generated by high-throughput sequencing technology in omics research, due to its ability to capture both the compositional structure of the data and overdispersion. However, an overlooked characteristic of the DM distribution is that it assigns positive probabilities to zero counts, even when the true probability of occurrence is zero. In this talk, I will introduce a novel Bayesian zero-inflated Dirichlet-multinomial (ZIDM) model for multivariate compositional count data with excess zeros, designed to improve scalability without sacrificing interpretability or imposing restrictive assumptions. I will then discuss extensions of the ZIDM model as a semiparametric mixture modeling framework that enables simultaneous clustering and automatic determination of the number of clusters in the data. The performance of the proposed methods will be evaluated through simulation and further illustrated using human microbiome data.