A decision theoretic approach to imputation in finite population sampling

Consider the situation where observations are missing at random from a simple random sample drawn from a finite population, In certain cases it is of interest to create a full set of sample values such that inferences based on the full set will have the stated frequentist properties even though the statistician making those inferences is unaware that some of lie observations were missing in the original sample. This article gives a Bayesian decision theoretic solution to this problem when one is primarily interested in making inferences about the population mean.