Double-mixing semiparametric logistic regression with unknown sizes

Abstract

Binomial data with unknown sizes often appear in biological and medical sciences and are usually overdispersed. All previous methods used parametric models and only considered overdispersion due to the variation of sizes. The proposed semiparametric model considers overdispersion due to the variation of sizes and that of probabilities. By doing this, it can include variations caused by observations, missing covariates, and random measurement errors in covariates. An Expectation Conditional Maximization algorithm is provided to stabilize the loglikelihood optimization. Selecting the number of support points of the mixing distributions and the bootstrap methods are also discussed. Simulation is done to evaluate the performance of the proposed model. Two real examples are used to illustrate the proposed model.