Assessing differences of mean estimates from longitudinal surveys, with an application to Research and Development Survey

• Activity Number: 352 - Small Area Estimation, Analysis of Complex Sample Survey Data, and New Advances for Health Surveys
• Type: Contributed
• Date/Time: Thursday, August 12, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Survey Research Methods Section
• Author(s): Rong Wei* and Yulei He and Van Parsons and Paul Scanlon
• Companies: National Center for Health Statistics, CDC and US Centers for Disease Control and Prevention and CDC and National Center for Health Statistics, CDC
• Keywords: US population health; RANDS; COVID-19 Pandemic; on-line surveys; longitudinal design; partial overlapping samples

Abstract

The Research and Development Survey (RANDS) is based on a web-panel platform designed for conducting questionnaire evaluation and statistical research and conducted by the National Center for Health Statistics (NCHS). During the COVID-19 pandemic in 2020, the survey added a component, RANDS during COVID-19, to focus on the general population’s health experience. The first two rounds of the survey were designed to be longitudinal so that data were collected repeatedly from the same set of subjects. In this study we propose statistical inference methods to assess differences of RANDS during COVID-19 estimates between the two periods. We consider two approaches: one is an extension of the two sample t-test and the other is a regression method. Both methods aim to account for the correlations between paired subjects as well as those induced by shared PSUs in the survey. These methods are demonstrated using variables such as self-rated health status, health insurance coverage, and outcomes of anxiety and depression.