A new Multivariate Methodology to Capture Spatial and Temporal Variability in Data: Application to Breast Cancer Mortality in Ecuador using Polylithogenic Offsets

Abstract

In the last decade, breast cancer mortality in Ecuador has increased considerably, with significant variability between the northern and southern regions of the country. Some provinces have significantly higher mortality rates, which are not evident when analyzing the spatial and temporal dimensions separately. The objective is to investigate breast cancer mortality in Ecuador using a novel multivariate statistical technique, the Time-Spatial Weighted Logistic Biplot (TSWLB), which has broad applicability in various fields. TSWLB integrates Geographically Weighted Principal Component Analysis (GWP-CA), the Mann-Kendall Nonparametric Statistical Test, and the External Logistic Biplot, allowing for the simultaneous representation of the spatial and temporal dimensions. The applied analysis classifies provinces into four distinct groups, each with distinct patterns and trends. Priority regions with a statistically significant increase in breast cancer mortality in recent years were identified. Furthermore, the Plithogenic Offsets approach is incorporated to model uncertainty and indeterminacy in the data, prioritizing regions by considering contradictory, non-contradictory, and indeterminate attributes, strengthening the identification of critical areas for intervention. These findings offer crucial information for policy makers to optimize resource allocation and develop effective prevention and treatment strategies in the most affected areas.