Abstract

Handwriting analysis generates continuous measurements that capture variations in geometric and spatial characteristics. Although these measurements provide quantitative descriptions of handwriting pattern, their multivariate associations may be difficult to interpret directly, particularly when several features are considered simultaneously. Transforming continuous measurements into categorical representation offers an alternative approach for exploring these associations through Multiple Correspondence Analysis (MCA). However, the choice of categorization strategy may influence the resulting association structure and its low dimensional visualization. This study investigates the transition from continuous handwriting measurements to categorical association maps, with particular attention to the sensitivity and interpretability of the resulting representation. Five image-derived handwriting features are considered, including letter height, consistency, letter spacing, word spacing, and stroke density. Data-driven categorization strategies are examined to construct categorical feature representation, which are subsequently analyzed using Classical MCA and Adaptive Threshold MCA (AT-MCA). The analysis focuses on changes in category configurations, inertia distribution, and the stability of low-dimensional association maps under different categorization schemes. By examining how discretization choices affect the graphical representation of multivariate association, this study extends previous work on AT-MCA toward a broader investigation of representation sensitivity. The proposed analytical perspective highlights the importance of considering both the construction of categorical variables and the stability of their visual representation when interpreting complex multidimensional data. The findings are intended to inform the use of MCA-based visualization in handwriting analysis and other application involving continuous-to-categorical data transformation.