|
Abstract:
|
Many State and local communities would like to design data collection efforts that are comparable to the National Health and Nutrition Examination Survey. To assist these design efforts, stage-specific variance components for major health variables measured in NHANES need to be produced. In NHANES, the sampling for specific demographic domains creates an unbalanced design that, along with small sample sizes, impacts the ability to produce reliable estimates for some methods of estimation. In this presentation, alternative estimation procedures are examined, specifically a simple naive estimator, a classic design based estimator, and an estimator based on a random effect model. Estimated variance components are presented for variables such as cholesterol levels, Diabetes prevalence, as well as variables for environmental containments such as Biphenyl-A and BDE-47.
|
