Techniques for Fusing Survey Modules: Respondent Matching and Data Imputation

Abstract

Survey respondents want short, applicable surveys they enjoy. In order to continue to receive good data, we need to capture respondents’ input through the means and within the capacity they are willing to provide. One way to meet the respondents’ needs is to break the traditional 20 -30 minute survey into 5-10 minute modules. These shorter modules appeal to a wider audience and work well in a mobile environment, but have large amounts of missing data which can be problematic for advanced research techniques. We examine two possible techniques to fusing the modular surveys back into a data set analysts can use to perform advanced analytics: hot-deck data imputation and respondent matching. We demonstrate how each technique works in a control experiment with a 25 minute A&U (attitude and usage) study on frozen carbonated beverages. Lastly, we discuss the implications and opportunities for future research into these fusion techniques.