Noninformative nonparametric quantile estimation for simple random samples

For noninformative nonparametric estimation of finite population quantiles under simple random sampling, estimation based on the Polya posterior is similar to estimation based on the Bayesian approach developed by Ericson (1969, JRSSB, 31, 195-233) in that the Polya posterior distribution is the limit of Ericson’s posterior distributions as the weight placed on the prior distribution diminishes. Furthermore, Polya posterior quantile estimates can be shown to be admissible under certain conditions. We demonstrate the admissibility of the sample median as an estimate of the population median under such a set of conditions. As with Ericson’s Bayesian approach, Polya posterior based interval estimates for population quantiles are asymptotically equivalent to the interval estimates obtained from standard frequentist approaches. In addition, for small to moderate populations, Polya posterior based interval estimates for quantiles of a continuous characteristic of interest tend to agree with the standard frequentist interval estimates.