ABSTRACT
Chapters 5 through 10 presented design-based approaches to the most common statistical analyses of complex sample survey data. Models certainly played an important role in defining these analyses, but in each case, the estimation and inferences for population parameters were based on the final survey weights (incorporating unequal probabilities of selection into the sample, nonresponse adjustments, and poststratification adjustments) and the large sample properties associated with the expected sampling distribution of estimates under repeated sampling from the chosen probability sample design. As described in Section 2.2, the “tried and true” design-based techniques for the analysis of complex sample survey data are appealing because they require minimal assumptions, produce consistent estimates of the target population parameters, and provide robust measures of the sampling variability of the estimates.
