A distance based test on random trees

TL;DR

This paper introduces a simple statistical test based on distances between empirical mean trees to compare populations of trees, demonstrating its effectiveness in distinguishing different means through simulations and real genomics data.

Contribution

It proposes a novel distance-based test for comparing tree populations, extending the analogy of the two-sample z test to tree-structured data.

Findings

Test effectively separates distributions with different means

Performance validated through simulations on Galton-Watson processes

Applied successfully to genomics data

Abstract

In this paper, we address the question of comparison between populations of trees. We study an statistical test based on the distance between empirical mean trees, as an analog of the two sample z statistic for comparing two means. Despite its simplicity, we can report that the test is quite powerful to separate distributions with different means but it does not distinguish between different populations with the same mean, a more complicated test should be applied in that setting. The performance of the test is studied via simulations on Galton-Watson branching processes. We also show an application to a real data problem in genomics.