Ukuran kesesuaian analisis korespondensi melalui analisis procrustes

Abstract

Correspondence analysis is a part of multivariate analysis studying the relationship of two or more variables which are displayed in rows and columns simultaneously from contingency table in low dimensional space using the Chi-square distance. From correspondence analysis, it is obtained row and column profiles co-ordinate matrix. This study aims to calculate the goodness-of-fit of correspondence analysis via Procrustes analysis and applied to two examples of data, i.e. the relationships between categories of smokers and groups of employees and the relationship between the province and the main jobs. To obtain the Procrustes analysis, we need to determine minimum difference through three geometric transformations, i.e. translation, rotation, and dilation. In correspondence analysis, we need to do three transformations on Procrustes analysis to obtain goodness-of-fit in row and column profiles co-ordinate matrix. The result of Procrustes analysis for relationship between employee groups and smoking habits to row and column profiles is 98.64% and 99.53% respectively. While the result of Procrustes analysis for relationship between province and the main jobs to row and column profiles is 91.83% and 88.25% respectively.