# ANOVA for Data in Metric Spaces, with Applications to Spatial Point Patterns

**Authors:** Raoul Müller, Dominic Schuhmacher, Jorge Mateu  
**arXiv:** [2201.08664](https://arxiv.org/abs/2201.08664)  
**Date:** 2022-01-21  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2201.08664v2/

## Key points

- Study focuses on ANOVA-like procedures for testing group differences in a metric space
- Introduces new procedure based on Levene's test to detect differences in dispersion using pairwise distances
- Advantageous when computing barycenters or generalized means is slow or infeasible
- Application of procedures to spatial point pattern data reveals significant influences of factors within the data
- Translation of point patterns into two-dimensional space through classical multidimensional scaling (mds) based on TT-distances for visualization and comparison
- Methodology successfully identifies subtle but important differences in real-world scenario involving bubbles in mineral flotation process

## Summary

In this study, we delve deeper into the analysis of ANOVA-like procedures for testing group differences based on data in a metric space. We introduce a new procedure that builds upon the classic Levene's test and focuses on detecting differences in dispersion using pairwise distances of data points. This approach proves to be advantageous in situations where computing barycenters or generalized means is slow or even infeasible. To provide more context, we explore the application of these procedures to spatial point pattern data. By examining the numbers of points per pattern, we are able to detect significant influences of factors within the data. Through classical multidimensional scaling (mds), we translate point patterns into a two-dimensional space based on TT-distances, allowing us to effectively visualize and compare patterns. Furthermore, through simulation studies and application to a real-world scenario involving bubbles in a mineral flotation process, our methodology successfully identifies subtle but important differences that impact our statistical tests. Overall, our refined analysis not only showcases the effectiveness of our new procedure but also highlights its practical applications in complex datasets such as spatial point patterns. , , , , .