In this paper, an approach to hypothesis testing for two populations of unlabelled networks is proposed. The Graph Space framework is integrated in order to handle networks that differ in the number of nodes or in node labelling. Permutation tests are considered, along with flexible choices on inter-point test statistics based on intrinsic metrics. Then, distances between networks are computed relying on approximate matching algorithms, in order to find optimal graphs alignments. Simulation studies show effectiveness and superior performance of the method with respect to alternative approaches. Finally, usefulness of the method is shown with an application to football players' passing networks. The proposed approach is available in the R package nevada.
In questo articolo, si propone un approccio al test d'ipotesi per reti non etichettate. Per trattare reti che differiscono nel numero o nelle etichette dei nodi, si considera la teoria sullo Spazio dei Grafi. Si adottano test di permutazione, garantendo flessibilità nella scelta di statistiche test "inter-point" basate su metriche intrinseche. Dunque, si valutano le distanze tra reti, affidandosi ad algoritmi di matching approssimato, in modo da ottenere allineamenti ottimali dei grafi. Le simulazioni rivelano efficacia e prestazioni superiori del metodo rispetto ad approcci alternativi. Infine, si mostra l'utilità del metodo nell'applicazione a reti di passaggio tra calciatori. L'approccio proposto è disponibile nel pacchetto R nevada.
Two-population test for unlabelled networks
MULÈ, DANIELE
2021/2022
Abstract
In this paper, an approach to hypothesis testing for two populations of unlabelled networks is proposed. The Graph Space framework is integrated in order to handle networks that differ in the number of nodes or in node labelling. Permutation tests are considered, along with flexible choices on inter-point test statistics based on intrinsic metrics. Then, distances between networks are computed relying on approximate matching algorithms, in order to find optimal graphs alignments. Simulation studies show effectiveness and superior performance of the method with respect to alternative approaches. Finally, usefulness of the method is shown with an application to football players' passing networks. The proposed approach is available in the R package nevada.File | Dimensione | Formato | |
---|---|---|---|
2022_07_Mulè_01.pdf
accessibile in internet per tutti
Descrizione: Thesis
Dimensione
1.06 MB
Formato
Adobe PDF
|
1.06 MB | Adobe PDF | Visualizza/Apri |
2022_07_Mulè_02.pdf
accessibile in internet per tutti
Descrizione: Executive Summary
Dimensione
338.06 kB
Formato
Adobe PDF
|
338.06 kB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/190775