A new method for multispace analysis of multidimensional social exclusion.

Area of Science:

• Social Sciences
• Geographic Information Science
• Statistical Analysis

Background:

• Social phenomena are complex and geographically dependent.
• Composite indicators, often using Principal Component Analysis (PCA), represent these phenomena but suffer from data sensitivity and limitations in multi-space-time comparisons.
• Existing methods like PCA can lead to informational loss and hinder comparative analyses across different geographic scales.

Purpose of the Study:

• To introduce Robust Multispace PCA, a novel method designed to overcome the limitations of traditional PCA for analyzing multidimensional social phenomena.
• To enhance the accuracy and informativeness of composite indicators for social exclusion.
• To facilitate more robust comparisons of social phenomena across multiple geographic spaces and time points.

Main Methods:

• Weighted sub-indicator importance based on conceptual relevance.
• Non-compensatory aggregation to ensure weight significance.
• Dimension aggregation to balance weight structures.
• A novel outlier-eliminating scale transformation function for multispatial comparison.

Main Results:

• The Robust Multispace-PCA method significantly reduces informational loss by 1.52 times compared to traditional methods.
• Demonstrated improved accuracy in composite indicators for social exclusion in urban areas across eight cities.
• The new scale transformation effectively handles outliers, enabling reliable multispatial comparisons.

Conclusions:

• Robust Multispace-PCA offers a more informative and accurate representation of multidimensional social phenomena.
• The method's ease of use makes it suitable for researchers and policymakers.
• It supports the development of targeted policies across multiple geographic scales, enhancing evidence-based decision-making.