Simultaneous Object and Category Score Estimation in Joint Correspondence Analysis.

Area of Science:

• Multivariate Statistics
• Data Visualization
• Categorical Data Analysis

Background:

• Joint Correspondence Analysis (JCA) is a statistical technique for dimensionality reduction of multivariate categorical data, serving as an alternative to Multiple Correspondence Analysis (MCA).
• Current JCA methods lack simultaneous representation of object and category scores on visualization maps, hindering result interpretability.
• MCA suffers from an inherent underestimated variance problem.

Purpose of the Study:

• To propose a novel simultaneous object and category score estimation method for JCA.
• To address the underestimated variance issue present in MCA.
• To enhance the interpretability of JCA results through improved visualization and factor-analytic interpretation.

Main Methods:

• Developed a JCA parameter estimation method minimizing discrepancies between observed data and the JCA data model.
• This approach contrasts with existing methods relying on the JCA covariance model.
• Explored both geometric and factor-analytic interpretations of JCA solutions.

Main Results:

• The proposed method enables the joint representation of objects and categories on JCA maps, overcoming previous limitations.
• The new estimation technique addresses the underestimated variance problem.
• Demonstrated the utility and interpretability of the proposed JCA method through two real data analysis examples.

Conclusions:

• The novel simultaneous score estimation method significantly enhances the interpretability of JCA.
• This approach provides a more comprehensive understanding of inter- and intra-relationships within categorical data.
• JCA, with this enhancement, offers a powerful tool comparable to exploratory factor analysis for data exploration.