Adaptive Cluster Sampling

Abstract

One of the main methods of adaptive sampling is adaptive cluster sampling. As it involves unequal probability of sampling, standard Horvitz-Thompson and Hansen-Hurwitz estimators can be modified to provide unbiased estimates of finite population parameters along with unbiased variance estimators. These estimators are compared with each other and with conventional estimators. Confidence intervals are discussed, including bootstrap and empirical likelihood methods, and a biased estimator that we call Hajek’s estimator is described because of its link with this topic. The chapter closes with some theory about selecting networks without replacement.