|
Abstract:
|
Diabetes is a leading cause of death in the U.S. The National Health and Nutrition Examination Survey (NHANES) is ideal for estimating the prevalence of diabetes because it includes interview and laboratory data. However, the complex nature of NHANES can make selection of the appropriate sample weight challenging. To illustrate how to select the appropriate sample weight, the three prevalence measures in Health, United States (HUS)-total, diagnosed, and undiagnosed diabetes -were used. The appropriate sample weight depends on the analytic objective and data components used. The sample weighting options considered were: 1. Using three weights specific to the three prevalence measures: the interview weight for diagnosed diabetes, the fasting subsample laboratory weight for undiagnosed diabetes, and replicate weights for total diabetes; 2. Creating a new combined interview and fasting subsample laboratory weight for all three prevalence measures; 3. Using the fasting subsample laboratory weight for all three prevalence measures. This presentation will compare these three approaches' strengths and weakness.
|
