Inference on the marginal distribution of clustered data with informative cluster size (Q2442679)

The authors consider models for clustered data where the cluster size could carry information about the marginal distribution of the outcome. In the first sections the concept of informative cluster size is developed--formal definitions are given, examples illustrate the defined models. Based on a nonparametric approach quantities of interest are defined as functionals, and the properties of their sample counterparts are discussed. The following quantities are considered: Mean and variance functionals, sign and rank based functionals accompanied with testing procedures, and regression functionals. The theory is illustrated with an example on patients with incomplete spinal cord injuries.