Date/Time Date(s) - 26/03/202610:30 am - 11:30 am
Speaker: Tobias Kenney (Dalhousie University)
Location: Hamilton Hall, Room 410 & Zoom Zoom link: https://mcmaster.zoom.us/j/98308755554 (Passcode: MathBio)
Title: Ornstein-Uhlenbek models of microbiome data with measurement error. (Joint work with H. Gu and S. Li.)
Abstract: The Ornstein-Uhlenbek (OU) process is a simple, widely used model for processes with stochastic perturbations and mean-reversion. It has previously been applied to temporal microbiome data. However, microbiome data are subject to substantial measurement error, which can severely impact parameter estimation if not properly accounted for. Measurement error can be accounted for using a state-space model, where the actual unknown microbial abundance is modelled as following an OU process, and the observed abundance follows some measurement error distribution centred on the unknown observed value.
In this talk, I will discuss three projects in this research direction. The first inferrence and model adequacy testing for a 1-dimensional OU state-space model. The second is the sampling strategy for estimating the mean-reversion rate. Changing the time interval between samples can have a very large effect on the accuracy of the estimated parameters. The third is a multidimensional process. A typical microbiome includes thousands of different genera of microbes. The interactions between these genera are often of greater interest than the dynamics of individual genera. We therefore seek to fit a multidimensional OU process that incorporates these interactions.
Dr. Kenney is an Assistant Professor in Dalhousie’s Department of Mathematics and Statistics with research interests in a wide variety of topics, mostly related to algebra, logic, combinatorics, and statistical genetics and phylogeny.