Comparing Methods for Testing Association in Tables with Zero Cell Counts Using Logistic Regression

Autor: Nurin Dureh, Hilary Green, Chamnein Choonpradub
Rok vydání: 2016
Předmět:
Zdroj: Proceedings of the International Conference on Computing, Mathematics and Statistics (iCMS 2015) ISBN: 9789811027703
DOI: 10.1007/978-981-10-2772-7_13
Popis: Logistic regression is one of the most useful methods used to describe the relationship between a binary dependent variable and a set of independent variables. However, when any of the counts are zero, a nonconvergence problem will occur. A procedure for solving such a problem has been proposed by Firth (Biometrika 80:27–38, 1993 [4]), and provides finite parameter estimates based on penalized maximum likelihood. This study suggests a simpler method which involves modifying the data by replacing the zero count by one and doubling the corresponding nonzero count. Results show that this simple data modification method gives similar results to those from the Firth’s procedure.
Databáze: OpenAIRE