Abstract:
|
Survey nonresponse has been increasing in the National Health and Nutrition Examination Survey (NHANES). Adjusting sample weights for nonresponse may reduce the bias of health estimates. However, as survey responses for non-respondents and the reasons for non-responding are not available, both the magnitude of nonresponse bias and whether sample weight adjustment reduce bias are unknown. NHANES collects participants' information in three stages: screening at home, interviewing at home, and examination in a mobile examination center (MEC). The screening stage determines survey eligibility. Demographic, socioeconomic, and health-related information is obtained in the home interview. In this simulation study, nonresponse models using interview information are used to create informative missingness. Using the simulated nonresponse data, sample weights with nonresponse adjustments based on the screening information are created. To evaluate whether sample weight adjustments reduce bias, the weighted health estimates from the simulated data are compared to those from the whole sample.
|