Quantitative habitat models of wildlife-habitat relationships are developed to formalize our current understanding about an ecological system. A habitat association model is one of these models that is useful for answering questions about how the habitat is occupied, how much growth habitat is required by the animal, or how the animal selects its food and habitat. Radio telemetry is adopted as a technique for studying home range and habitat use. The major objective of a radio telemetry study is to collect behavioral or demographic data in order to be able to estimate population parameters for home range and habitat selection. A radio telemetry study is a kind of multinomial experiment. The Logistic Regression Model is often used for estimating the relationship between animal activities and the habitat characteristics of the location used (animal preference). However, this model is not a good model for the telemetry data. Under this model, the slope parameter estimate becomes lower and farther from the true value as the Average Habitat Quality (AHQ) increases, with Diversity fixed. The Multinomial Model is better suited to telemetry data. Using the Logistic Regression Model, a habitat association study can be conducted in conjunction with adaptive cluster sampling. In terms of the variance of the regression parameter estimate, adaptive cluster sampling is better than simple random sampling. Adaptive sampling plans are also satisfied for habitat association analysis with imperfect detectability. / Graduation date: 1995

Multivariate statistical techniques were applied to the development of habitat suitability models for the Mt. Graham red squirrel, an endangered species. A digital map data base and a geographic information system (GIS) were used to support the analysis and provide input for two logistic multiple regression models. Squirrel presence/absence is the dichotomous dependent variable whose probability the models pretend to predict. Independent variables are a set of environmental factors in the first model, and locational variables in the second case, where a logistic trend surface was developed. Bayesian statistics were then used to integrate the models into a combined model. Potential habitat losses resulting from the development of an astronomical observatory were assessed using the environmental model and are found to represent about 3% of currently available habitat.