[sci.math.stat] Fit Assessment & Identification of Functional Form in Logistic Regression

ruth@utstat.uucp (Ruth Croxford) (10/28/88)

Anyone who is interested is invited to attend the next Biostatistics Rounds,
Wednesday, November 9, 5:00 p.m. at the Princess Margaret Hospital, room 689

Title:    Fit Assessment and Identification of Functional Form in Logistic
          Regression
Speaker:  Salomon Minkin, Ontario Cancer Institute

Summary:
For the analysis of binary data using logistic regression, there is a need
for simple techniques for detecting model inadequacies and suggesting 
corrective actions.  This paper proposes an approach based on the use of
line segments to approximate the effect of a predictor.  This provides the
basis for easily implemented goodness of fit tests and graphical methods 
for identifying the nonlinearities.  Methods for selecting the join points
are discussed, and their use is illustrated with data from breast cancer
patients.