Modelling Data That Exhibit an Excess Number of Zeros: Zero-Inflated Models and Generic Mixture Models

摘要: In biomedical research, data generated as a consequence of the count process can often possess an ‘excess’ zeros (e.g. geographical incidence rates, hospital death rates). Whilst there are strategies for analysing such data, some be biased where underlying generation is not carefully considered. This exacerbated also multilevel, since hierarchical extensions to zero-inflated model do always satisfy assumptions. We therefore review modelling single-level and show why standard Poisson binomial models (i.e. one latent class has central location zero) require membership predicted by covariates in regression part model. introduce generic mixture reveal limitations their interpretation number circumstances. With nested or with excess zeros, upper-level distributional assumptions may upheld multilevel models, thereby requiring alternative strategies; Chap. 7 we illustrate semi-parametric solution this problem.