We propose to initiate analysis of the tendencies of respondents to the Health and Retirement Study (HRS) for nonresponse and response errors in answering related questions, as well as across the instrument and between waves.
We aim to learn whether certain types of respondents tend to systematically provide inaccurate or incomplete information. It has been common in analysis of data quality problems in the MRS and other surveys to study responses to a particular item of interest. However, respondents may have tendencies to provide/not provide accurate/inaccurate information across related questions, or perhaps across the entire instrument. We will investigate how nonresponse and response errors are related across questions, and we will examine the statistical association of response patterns with observable characteristics of respondents. It has been common in analysis of data quality problems to address nonresponse and response errors as separate matters. However, these two facets of data quality may be related. We intend to study the relationship between nonresponse and the accuracy of the data that respondents do provide. Skip sequencing is used in surveys to reduce respondent burden and the cost of interviewing. However, nonresponse and response errors to the opening questions used to determine skip sequences can substantially affect the quality of the data obtained on subsequent questions. We will study how skip sequencing in the HRS transmits nonresponse and response errors across survey items. We will investigate ways to cope with and mitigate the inferential problems arising from skip sequencing.
The Health and Retirement Study (HRS) is the premier source of publicly available information on the health and economic status of older Americans. This research will improve the ability of researchers to use the HRS while recognizing that some respondent records are incomplete and possibly error-ridden. ? ? ? ?
| Manski, Charles F; Molinari, Francesca (2010) Rounding Probabilistic Expectations in Surveys. J Bus Econ Stat 28:219-231 |
