On testing the hidden heterogeneity in negative binomial regression models (Q2312021)

To test the existence of hidden heterogeneity in negative binomial regression models, this paper develops score statistics under additive and multiplicative random effect models. Explicit forms of the test statistics and their asymptotic distribution are provided and the relationship between the test statistics derived from the two random effect models is discussed. A simulation study shows that the proposed approach leads to superior performance than existing methods in terms of controlling for type I error and power. Two real data analyses of number of articles produced in Ph.D. programs and of automotive claims are presented.