|
Abstract:
|
Iowa's State Board of Education carried out a stratified multistage sample survey to study the availability of employment preparation courses and the degree to which students enroll in those courses in Iowa's public schools. Three survey sampling issues arose in the design that was implemented. First, the design actually is not invariant. Usual variance estimators are compared to simulated sampling distributions. Second, some PSUs in the stratified multistage sampling design have only one PSU per stratum. The collapsed variance estimator, model-based estimators based on generalized variance functions, and a ratio estimator based on auxiliary variables are proposed and compared. Results are based on a simulation study. Third, as part of ongoing work, the relative costs of adding schools versus adding students to the sample will be evaluated.
|
