Abstract:
|
In observational studies, samples drawn from one or more study populations are used to make inferences concerning a target population. Such inferences require adjustment for population differences. Minimum discriminant information adjustment (MDIA) can be employed to weight samples to conform to known population information. In the case of simple random sampling, Haberman (1984) has derived large-sample properties for MDIA-based weighted sample means. In this paper, these results are generalized to complex sampling designs, including stratified sampling and two-stage sampling. To illustrate use of these new results and to evaluate accuracy of large-sample approximations, applications are made to a statewide program in Florida middle schools to provide reading coaching.
|