Use of random permutation model in rate estimation and standardization
Through integrating techniques from several areas in survey sampling, we develop an alternative method of deriving estimators using random permutation models under (stratified) simple random sampling without replacement. The finite random permutation model links the samples to the population. The joint permutation of response and auxiliary variables is modeled using seemingly unrelated regression. We use prediction theory from the super-population sampling literature to derive the linear unbiased minimum variance predictors of population means under the design-based framework using the finite estimating equation approach of Binder and Patak (1994). The predictors have functional forms similar to those derived using design-based, model-assisted and calibration approaches, but depend on neither superpopulation nor regression model assumptions. We applied the results to standardization of multiple rates, and illustrate how our methods account for the covariance of the standardized rates, unlike conventional standardization methods.
0573: Public health