Correction for the shrinkage effect in gaussian graphical models

• Victor Guryev ; Rainer Bischoff ; Péter Horvatovich ; Marco Grzegorczyk ^[1] ; Victor Bernal ^[1]
  1. [1] University of Groningen

     University of Groningen

     Países Bajos

     
• Localización: Proceedings of the 35th International Workshop on Statistical Modelling : July 20-24, 2020 Bilbao, Basque Country, Spain / Itziar Irigoien Garbizu (ed. lit.), Dae-Jin Lee (ed. lit.), Joaquín Martínez Minaya (ed. lit.), María Xosé Rodríguez Álvarez (ed. lit.), 2020, ISBN 978-84-1319-267-3, págs. 290-293
• Idioma: inglés
• Enlaces
  • Texto Completo Libro
• Resumen

  • Gaussian graphical models (GGMs) are probabilistic graphical models based on partial correlation. A GGM consists of a network of nodes (representing the random variables) connected by edges (their partial correlation). To infer a GGM, the inverse of the covariance matrix (the precision matrix) is required. The main challenge is that when the number of variables is larger than the sample size, the (sample) covariance is ill conditioned (or not invertible). Shrinkage methods consist in regularizing the estimator of the covariance matrix to make it invertible (and well conditioned); however, the e ect of the shrinkage on the nal network topology has not been studied so far.