ABSTRACT
There are two basic approaches to statistical analysis from sample survey data. The first is the one used in the development in this book, where a statistical model for the sampled population is assumed and inference is carried out assuming that this population model holds. This is usually referred to as model-based analysis, and the model for the population is often referred to as a superpopulation model, because the values inBU = {DU ,RU , iU ,ZU} are a realization of the (multivariate) random variable whose characteristics are specified by the model. Inference is based on the density 〈Bobs〉 of the observed data 〈Bobs〉 = {Dobs,Rs, iU ,ZU}. If selection is noninformative given ZU , model-based analysis is based on the density of {Dobs,Rs,ZU}. This in turn reduces to {Dobs,ZU} if there is complete response or the nonresponse is noninformative given ZU .
