Data Quality in the UAS and the Flag File

The UAS uses multiple approaches to ensure high quality data. Surveys include checks for implausible answers and inconsistencies in the programmed questionnaires. UAS researchers make generous use of randomizations, which both serve to detect possible order effects as well as straightlining. In addition, we record question timings for each survey that may be used to detect speeding.

From time to time, the UAS helpdesk receives and rejects requests from individuals who are not qualified to join the panel because they live in a household that was not selected in our recruitment sample (e.g. a friend of a participant who does not live at the same address). To ensure adherence to our sampling protocol, UAS participants who request to add additional household members to the panel (members not included in the initial household enumeration) are required to provide evidence that the individual lives at the selected sample address. In addition, UAS staff regularly verify addresses to mail information and items to participants. If a participant address does not match the household address, an inquiry is made and action taken if necessary.

Very occasionally the UAS has removed participants, or participant data, from the UAS to protect data quality. One example is a participant who began helpfully doing surveys for themselves and their spouse when the spouse’s health began to decline. Another example was a participant who did not speak English or Spanish, so another household member had been translating for them. In cases like these which affect data quality, the UAS takes action to correct the situation, educates participants, and closes accounts as needed. When compromised data can be removed from the system, we do so. However, historical survey data files and data products may still include data contributed by these individuals. Thus, for situations where participants have been removed from the study, we have created the Data Flag File. Researchers working with existing UAS data are encouraged to use the Flag File to identify any historical data that has been flagged for removal.

The Flag File consists of four variables: uasid; status (1= soft flag, still under evaluation / 2 = hard flag, remove data); begin and end which provide dates corresponding to the period in which the problem occurred. If no dates are provided for a status value of 2, this indicates that the respondent's data should be ignored in its entirety; if only a start date is provided, all respondent's data on or after that date should be ignored. This allows investigators to identify surveys that were and were not affected.

Download Flag File: