Understanding the Autologistic Model
Document Type
Presentation Abstract
Presentation Date
4-23-2012
Abstract
For a binary response variable the logistic model is commonly implemented to describe the probability of success as a function of one or more covariates. As long as the response variables are independent, such a paradigm is appropriate. However, when binary responses on a regular lattice are observed in space, spatial dependencies typically exist and the logistic model is rendered invalid. The autologistic model is an intuitive extension of the logistic model that accommodates such a lack of spatial independence. Unfortunately, the normalizing constant for its joint likelihood function is usually intractable, and, therefore, methods other than maximum likelihood are needed to estimate the parameters of the autologistic model.
In this talk we will review the logistic model, introduce the autologistic model, and present four methods of parameter estimation for the autologistic model, including pseudolikelihood, Markov chain Monte Carlo maximum likelihood, generalized pseudolikelihood, and block generalized pseudolikelihood. A fire occurrence data set from Oregon and Washington will be used throughout the presentation as a means to motivate and illustrate the aforementioned concepts. As this talk is a prelude to a forthcoming defense, various topics comprising the defense presentation will also be previewed.
Recommended Citation
Purdy, Jordan Earl, "Understanding the Autologistic Model" (2012). Colloquia of the Department of Mathematical Sciences. 397.
https://scholarworks.umt.edu/mathcolloquia/397
Additional Details
Monday, 23 April 2012
3:10 p.m. in Math 103
4:00 p.m. Refreshments in Math Lounge 109