Discrete beta-type models

Abstract

A more interpretable parameterization of a beta density is the starting point to propose an analogous discrete beta (d:b:) distribution assuming values on a finite set. Thus a smooth estimator using d:b: kernels is considered. By construction, it is both well-defined and free of boundary bias. Taking advantage of the discrete nature of the data, a technique of smoothing parameter selection is also proposed in moderate-to-large samples. Finally, a real data set is analyzed in order to appreciate the advantages of this nonparametric proposal. © Springer-Verlag Berlin Heidelberg 2010.