A computationally efficient estimator for large clustered non-Gaussian data

• Alvaro J. Alvaro J. Flórez ^[1] ; Geert Molenberghs ; Geert Verbeke ; Pavlos Mamouris ^[2] ; Bert Vaes ^[3]
  1. [1] University of Hasselt

     University of Hasselt

     Arrondissement Hasselt, Bélgica

     
  2. [2] KU Leuven

     KU Leuven

     Arrondissement Leuven, Bélgica

     
  3. [3] Academisch Centrum voor Huisartsgeneeskunde

  Mostrar afiliaciones +
• 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. 79-84
• Idioma: inglés
• Enlaces
  • Texto Completo Libro
• Resumen

  • The generalized linear mixed model (GLMM) is one of the most frequently used techniques to analyze clustered non-Gaussian data. Commonly, the GLMM is fitted by maximizing the marginal (log-)likelihood, i.e., integrating out the random effects. However, this whole maximisation may require a considerable amount of computing resources. Although computationally manageable with medium to large data, it can be too time-consuming or computationally intractable with very large clusters and/or with a large number of clusters. To overcome this, a fast two-stage estimator for correlated non-Gaussian data is presented. It is rooted in the pseudo-likelihood split-sample methodology. Based on simulations, it shows good statistical properties, and it is computationally much faster than full maximum likelihood. The approach is illustrated using a large dataset belonging to a network of Belgian general practices