The PC algorithm is the most known constraint-based algorithm for selecting a directed acyclic graph. For Gaussian distributions, it infers the structure using conditional independence tests based on Pearson correlation coefficients. The Rank PC algorithm, based on Spearman correlation, has been recently proposed when data are drawn from a Gaussian Copula model. We propose a modified version of the Grow-Shrink algorithm, based on the recovery of the Markov blanket of the nodes and on the Spearman correlation. In simulations, our Copula Grow-Shrink algorithm performs better than PC and Rank PC ones, according to structural Hamming distance.

A constraint-based algorithm for nonparanormal data / Musella, Flamina; Vicard, Paola; Vitale, Vincenzina. - (2017). (Intervento presentato al convegno CLADAG 2017 - 11th Meeting of the Classification and Data Analysis tenutosi a Milano).

A constraint-based algorithm for nonparanormal data

Vicard Paola;Vitale Vincenzina
2017

Abstract

The PC algorithm is the most known constraint-based algorithm for selecting a directed acyclic graph. For Gaussian distributions, it infers the structure using conditional independence tests based on Pearson correlation coefficients. The Rank PC algorithm, based on Spearman correlation, has been recently proposed when data are drawn from a Gaussian Copula model. We propose a modified version of the Grow-Shrink algorithm, based on the recovery of the Markov blanket of the nodes and on the Spearman correlation. In simulations, our Copula Grow-Shrink algorithm performs better than PC and Rank PC ones, according to structural Hamming distance.
2017
CLADAG 2017 - 11th Meeting of the Classification and Data Analysis
joint normal copula; rank-based correlation; Grow-Shrink algorithm
04 Pubblicazione in atti di convegno::04b Atto di convegno in volume
A constraint-based algorithm for nonparanormal data / Musella, Flamina; Vicard, Paola; Vitale, Vincenzina. - (2017). (Intervento presentato al convegno CLADAG 2017 - 11th Meeting of the Classification and Data Analysis tenutosi a Milano).
File allegati a questo prodotto
File Dimensione Formato  
A CONSTRAINT-BASED ALGORITHM FOR NONPARANORMAL DATA.pdf

solo gestori archivio

Dimensione 131.33 kB
Formato Adobe PDF
131.33 kB Adobe PDF   Contatta l'autore

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11573/1411420
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact