Methods of Discrimination Among Stochastic Models of the Negative Binomial Distribution with an Application to Medical Statistics

The Negative Binomial Distribution (NBD) can be derived from several underlying stochastic processes. For the NBD to be useful in practice as a descriptive model, it would be of interest to know which underlying process it is characterizing. In this paper, methods for discriminating among three different NBD underlying processes, proneness, linear contagion, and spells, are developed and studied. The methods are based on the characteristics of the processes and employ the theory of the bivariate NBD and waiting times. The NBD has been shown to consistently fit data on the incidence of disease. Knowledge of the underlying process generating the data is important for meaningful interpretation of such findings.

Duke Scholars

Author Frank Wesley Rockhold Biostatistics & Bioinformatics, Division of Biostatistics