SUMMARY The polya posterior gives a noninformative Bayesian justification for various single-stage sampling procedures. Here this type of reasoning is extended to two-stage cluster sampling problems. The frequency properties of the resulting procedures are studied.
