ABSTRACT
Probability sampling is a recent innovation in the history of statistics. Many investigators take samples that are not chosen using random selection methods, called nonprobability samples. Examples include administrative records, quota samples, judgment samples, and convenience samples. The main concern with nonprobability samples, though, is that estimates calculated from them may be biased for the corresponding population quantities. This chapter presents a statistical framework for considering the bias of estimates from nonprobability samples and presents methods that have been proposed to estimate population quantities from samples that have not been selected randomly. It concludes with a comparison of the advantages and disadvantages of nonprobability samples when compared with low-response-rate probability samples.
