Methods for Analysis of Recall of Fires in a Retrospective Survey

• Activity Number: 519
• Type: Contributed
• Date/Time: Wednesday, August 5, 2009 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Michael A. Greene*+ and Craig D. Andres
• Companies: U.S. Consumer Product Safety Commission and U.S. Army Aberdeen Test Center
• Keywords: adverse events ; recall delay ; imputation

Abstract

The U.S. Consumer Product Safety Commission staff conducted a national probability survey on residential fires, including fires not attended by fire departments. Information was collected from 916 households, who told interviewers they had one or more fires in the past 90 days. Inability to remember complete dates of fires and decreasing estimates of annual fires with longer recall periods suggested that respondents were unable to recall some fire incidents. We adapt a method by Warner, Schenker, Heinen and Fingerhut (2005), for selecting the length of the recall period to minimize the mean square error (MSE) of the estimated number of fires, where the MSE is calculated from the recall bias and the sampling and imputation variance. The results suggest that respondents do not recall events for long, which has cost and design implications for retrospective surveys.