Simulation Approach for Determining Use of Mahalanobis Distance to Reduce Nonresponse Bias

• Activity Number: 695
• Type: Contributed
• Date/Time: Thursday, August 8, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Survey Research Methods Section
• Author(s): Jennifer Cooney*+ and Peter Siegel and Melissa Cominole and Bryan Shepherd
• Companies: RTI International and RTI International and RTI International and RTI International
• Keywords: Mahalanobis Distance ; Nonresponse Bias ; Simulation

Abstract

Objective: Develop a simulation which determined if nonresponse bias for the 2008/09 Baccalaureate and Beyond Longitudinal Study can be reduced more effectively by targeting nonrespondents (NR) who are the most different from respondents (R).

Methods: The Mahalanobis Distance (MD) value was calculated for each sample member using respondent information after three months of data collection as the reference group. The distribution of the NR MD values was examined to set a cutpoint above which NRs were considered the most different from Rs.

NRs were reclassified as Rs under five scenarios with different conversion rates but with a fixed overall response rate. A base rate assumed equal conversion rates while the other sampling rates focused on increasing conversion rates above the cutpoint. New MD values were calculated for the remaining NRs to determine if there was a reduction in the bias distribution for each of the strategies.

Results: The simulation showed that larger conversion rates above the cutpoint reduced bias until XX% at which point rates beyond that resulted in higher bias due to focus on extreme responses.