Conditional maximum likelihood estimation is widely used in epidemiological practice for log odds ratio regression problems. Andersen (1970) discusses conditions under which the conditional maximum likelihood estimator is consistent and asymptotically normal in settings where no covariates are present. This paper shows that an application of results of Fahrmeir and Kaufmann (1985) yields simple conditions under which the conditional MLE of the odds ratio regression coefficients is consistent and asymptotically normal.
- Conditional maximum likelihood estimator
- log odds ratio regression
- sparse strata asymptotics