Statistical modelling for complex binary data in biometrics
by
Tony Pettitt
Queensland University of Technology

Complex binary data arise variously in biometrics. Cases range from diseased or not diseased patients, present or absent plants, signals that are off or on. Measurements may be arranged in space, ordered by time or clustered in units. This talk considers a general model for binary data, a quadratic exponential family model, which is known as the autologistic model when specialized for spatial data. Like all statistical models, this family has its advantages and disadvantages and these will be described. A recent result of the speaker's which overcomes one difficulty, the computation of the normalising constant, will be described and various applications to data from agriculture, ecology and physiology will be given.

Date received: September 26, 2001