POWRE: Methology for Spatial Models for Binary Data

点击翻译按钮获取中文摘要

9806243 Hoeting The proposed research involves the development and application of new methodology for spatial models for binary data, particularly autologistic models. In the basic autologistic model, the response at a given site is predicted using a logistic function on the responses for the neighboring sites. For many applications of this model additional information is available, such as covariates that are related to the response of interest. This is the case when a sample of sites is collected over an area of interest. The autologistic model with covariates for sample data (Hoeting et al, 1997) allows the user to fully integrate all of this information into the model. The proposed research will develop new methodology to address four research problems related to the autologistic model with covariates including methodology for variable selection for the autologistic model with covariates, Bayesian model averaging (BMA) for the autologistic model with covariates, methodology to account for the probability of detection in sampling, and a space-time model for the autologistic model with covariates. Theses improvements will make the autologistic model with covariates useful for more complex problems.