BOOTSTRAPPING CORRESPONDENCE ANALYSIS

Abstract

Correspondence analysis, also known as homogeneity analysis is a statistical visualization method for graphically representing of two way contingency tables. Correspondence analysis is a method for exploring associations between sets of categorical variables and is one of the ordination techniques aimed at reducing data sets into a manageable number of variables. Relationships between categorical data can be analyzed with correspondence analysis and the results are obtained graphically in a two-dimensional space. Conversely to advantages of correspondence analysis, impact of any changes in the data cannot be determined by this analysis. At this point, bootstrapping correspondence analysis gives additional information by exhibiting ellipses of row and column profiles of the contingency table. This study applies bootstrap ellipses for graphical visualization of correspondence analysis on market research data of a baby food brand. The results show that with the bootstrapped ellipses, it is convenient to merge categories of the contingency table, resulting in dimension reduction. This result cannot be obtained with regular correspondence analysis.