Abstract
Lawrence R. Ernst (1995) "Maximizing and Minimizing Overlap of
Ultimate Sampling Units," Proceedings of the Section on Survey Research Methods,
American Statistical Association, 706-711.
Several general procedures have been developed for maximizing the expected number of
primary sampling units in common to the new and initial samples, without altering
unconditional selection probabilities, when redesigning a survey with a stratified,
multi-stage design. Previous procedures for the analogous problem for ultimate sampling
units (USUs) have been limited to simple cases, such as Poisson sampling. A procedure was
developed several years ago at the Bureau of Labor Statistics, and currently used in the
sample selection for the Occupational Compensation Surveys Program (OCSP), for increasing
overlap of USUs when both the initial and new samples are selected with equal probability
within a stratum. In this paper we present a modification of that procedure, which further
increases the overlap in certain situations. We then generalize this procedure to the
unequal probability case. Finally, we demonstrate how this approach can also be used when
it is desired to minimize, rather than maximize, the overlap of USUs. These procedures are
all computationally quite simple.
Last Modified Date: July 19, 2008
|