University of Washington
Department of Statistics

PACIFIC NORTHWEST STATISTICS MEETING
POSTER ABSTRACT


Thomas Lumley
Department of Biostatistics
University of Washington

Lisp-Stat tools for building GEE models

Friday, April 12, 1996.
1:00 P.M. HUB Auditorium

Abstract

The XLISP-Stat statistics system is convenient for implementing new statistical methods as it is freely available, reasonably fast and contains a powerful programming language and support for dynamic graphics. An object-oriented system for modelling longitudinal data using Generalised Estimating Equations has been developed in XLISP-Stat and will be presented. The system includes a wide range of choice for link function, error function and working correlation model. Design matrices can be constructed using a model formula specification for factor and metric variables and their interactions. Diagnostic plots are also available which make use of the dynamic graphics features of XLISP-Stat to link diagnostics for different variables and for different observations on an individual.