ABSTRACT
Clustered binary data have been analyzed mainly in a parametric way. The selection of the proper functional forms describing the dependence of all main and association parameters in a specific probability model is not always an easy task. There is clearly a need for flexible parametric models and, in case the design allows, for semi-and nonparametric approaches. This chapter illustrates how two popular classes of polynomial models, fractional and local polynomials, offer great flexibility in modeling clustered data.
