Stable computational methods for additive binomial models with application to adjusted risk differences

Abstract

Risk difference is an important measure of effect size in biostatistics, for both randomised and observational studies. The natural way to adjust risk differences for potential confounders is to use an additive binomial model, which is a binomial generalised linear model with an identity link function. However, implementations of the additive binomial model in commonly used statistical packages can fail to converge to the maximum likelihood estimate (MLE), necessitating the use of approximate methods involving misspecified or inflexible models. A novel computational method is proposed, which retains the additive binomial model but uses the multinomial–Poisson transformation to convert the problem into an equivalent additive Poisson fit. The method allows reliable computation of the MLE, as well as allowing for semi-parametric monotonic regression functions. The performance of the method is examined in simulations and it is used to analyse two datasets from clinical trials in acute myocardial infarction. Source code for implementing the method in R is provided as supplementary material (see Appendix A).