No abstract

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.. Biometrika Trust is collaborating with JSTOR to digitize, preserve and extend access to Biometrika. SUMMARY This paper presents a general framework for model-based statistics. The framework is based on extended concepts of sufficiency and ancillarity. The fundamental theorem of the new theory exhibits in a single equation the reference distributions for decision theory, model checking, and conditional inference for a general statistical model. The theorem includes the equations upon which Bayesian and frequentist statistics are founded. The general theory also fills two gaps previously missing from empirical Bayesian statistics: concepts of sufficiency and conditional inference.

We develop exact conditional methods for checking the fit of a logistic regression model to individual matched sets in case-control studies. Justifications are given for preferring this approach to conventional maximum likelihood-based methods. Data from two studies are used to illustrate the techniques.

Saddlepoint methods provide quick and easy approximations to significance levels for conditional tests of logistic regression parameters. We evaluate the accuracies of saddlepoint approximations for three well-known conditional tests: Bartlett's test for no three-factor interaction in a 2 x 2 x 2 table, the test for trend in a series of probabilities, and the exact test of no association in stratified 2 x 2 tables with a common odds ratio. General recommendations are suggested regarding the use of saddlepoint approximations for exact conditional significance levels.

No abstract