Adjusting for Survey Measurement Error with Accuracy Variables

• Activity Number: 88
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 4:00 PM to 5:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Damiao N Da Silva*+ and Chris Skinner
• Companies: S3RI/Universidade Federal do Rio Grande do Norte and London School of Economics
• Keywords: Survey sampling ; nonsampling bias ; cumulative distribution function ; pseudo maximum likelihood

Abstract

Many sample surveys gather information on the accuracy of the data collection process. For example, for questions related to salary in face--to--face interviews, the interviewer may record if the interviewee consulted the latest payment slip, an early payment slip or did not consulted any other source. Another example is when the interviewer records if a set of questions were answered very accurately, fairly accurately, not very accurately or not at all accurately. Although these types of data can be informative of the presence of measurement error, little is known how to use the accuracy data to produce corrected estimates of the parameters of interest. In this paper, we propose a methodology that incorporates accuracy variables and allows for adjusting for measurement error. Statistical properties of the estimators are examined through a simulation study. The results demonstrate the adjustments outperform the simple estimator that ignores the measurement error.