Additive and multiplicative statistical models in epidemiology invoke different assumptions regarding exposure-outcome relationships. While either may seem to provide an adequate fit in small data sets, a comparative parameter based on the correct set of assumptions will be more stable when applied in other contexts. Important implications of model selection and the scant literature on theoretical reasoning are presented. Given a lack of biological knowledge, a body of corroborative empirical knowledge would be helpful in choosing between statistical models. Traditionally, the fit of a multiplicative model is evaluated by a $ chi sp2$ test of homogeneity. However, this is a test rather than a measure and is sample-size dependent. In this thesis, the development of a new measure of heterogeneity of rate ratios, phi prime, is presented and applied to stomach cancer registry data. Results suggest homogeneity when comparing regions within countries, but some heterogeneity between continents.
Identifer | oai:union.ndltd.org:LACETR/oai:collectionscanada.gc.ca:QMM.56966 |
Date | January 1992 |
Creators | Scott, Susan C. (Susan Catherine) |
Contributors | Hanley, James A. (advisor) |
Publisher | McGill University |
Source Sets | Library and Archives Canada ETDs Repository / Centre d'archives des thèses électroniques de Bibliothèque et Archives Canada |
Language | English |
Detected Language | English |
Type | Electronic Thesis or Dissertation |
Format | application/pdf |
Coverage | Master of Science (Department of Epidemiology and Biostatistics.) |
Rights | All items in eScholarship@McGill are protected by copyright with all rights reserved unless otherwise indicated. |
Relation | alephsysno: 001323904, proquestno: AAIMM87702, Theses scanned by UMI/ProQuest. |
Page generated in 0.0116 seconds