|
Abstract:
|
The Bayesian approach to statistical inference has become popular in many areas of science, but the approach is not embraced widely for sample survey inference, where the design-based paradigm remains predominant. Reasons include concerns about the effects of model misspecification, the need for simple procedures for extensive survey analysis, and antipathy toward the subjective elements of Bayes when it comes to 'objective' reporting of survey results. The author argues that these objections can be overcome and the Bayesian paradigm is an excellent tool for survey inference, provided attention is paid to incorporating features of the survey design into the models and strong parametric assumptions are avoided when the sample size is large. Some applications of the Bayesian approach are offered in support of this claim.
|
