Goodness-of-Fit Testing of Log-Link Models for Categorical Outcome Data
We have contributed to recent advances in statistical methodology that have made relative risk estimation possible for follow-up data with continuous covariates. The mathematical models are the log binomial model for binary outcomes, and the log multinomial model for multinomial outcomes. The aim of this project is to investigate whether the fit of a log binomial model can be assessed using goodness-of-fit tests developed for the binary logistic model, and whether the fit of a log multinomial model can be assessed using goodness-of-fit tests developed for the multinomial logistic model.
Research Groups
Related Diseases
Staff
Team Leaders
External Collaborators
- Professor David Hosmer - University of Massachusetts