Fig. 1From: Odds ratios from logistic, geometric, Poisson, and negative binomial regression modelsComparison of information gains across count models. The figure shows FI(Y;θ)/FI(Z;θ) as a function of the p= Pr(Y>0) for various count models. As the probability of a positive count increases, the relative FI increases. The gains are largest for models with the most dispersion in the countsBack to article page