Effects of Misclassification in Sampling Stratification on Survey Estimates

• Activity Number: 475 - Recent Advances in Sample and Survey Design
• Type: Contributed
• Date/Time: Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Survey Research Methods Section
• Author(s): Tiandong Li* and Tracy Mattingly
• Companies: Health Resources and Services Administration and the Census Bureau
• Keywords: misclassification; frame quality; strata jumper

Abstract

In the 2018 National Sample Survey of Registered Nurses (NSSRN), the sample of Registered Nurses (RNs) was stratified by the status of being an Nurse Practitioner (NP) versus a non-NP, with NPs heavily oversampled. The stratum information in the sampling frame is not always consistent to the reported information in the interview. Such inconsistency may be due to recent transition from an RN to an NP and inability to identify the NP based on the frame information. When nurses in the non-NP stratum reported to be an NP, outlying large weights may appear in the NP analysis domain defined based on the interview data. Hence, the sample designed to meet analytic objectives could result in the loss of effective sample sizes in NP domains. In assessing the impact of this inconsistent information, the data obtained from the respondents is considered the true values. Using the NSSRN data, we evaluate the level of inconsistency between the two data sources and investigate the impact of such inconsistent information on survey estimates. We also explore the alternative methods in sampling and weighting to mitigate the impact.