Learning about Respondents’ Characteristics Using Standard Exploratory Data Analysis (EDA) Tools)

MoonJung Cho and Larry Lang

Abstract

With aid of computing power, EDA has made remarkable advancements especially in vi- sualization, clustering, and dimension reduction. We examined the characteristics of sur- vey non-respondents using standard EDA tools such as classical multidimensional scaling (MDS), spectral clustering, and topological networks. In addition, we applied classifica- tion and regression tree methods to identify the important variables which could better assess and interpret nonresponse rates. We applied these tools to analyze non-respondents’ characteristics at the initiation stage in the U.S. International Price Program Survey.