Visualising model selection in regression

  1. Cecilio Mar-Molinero 1
  2. Carlos Serrano-Cinca 2
  3. Fabiola Portillo 3
  1. 1 University of Kent, UK, and University Autonoma of Barcelona, Spain
  2. 2 Universidad de Zaragoza
    info

    Universidad de Zaragoza

    Zaragoza, España

    ROR https://ror.org/012a91z28

  3. 3 Universidad de La Rioja
    info

    Universidad de La Rioja

    Logroño, España

    ROR https://ror.org/0553yr311

Montrer des affiliations +
Actes de conférence:
4th Stochastic Modeling Techniques and Data Analysis International Conference with 5th Demographics Workshop. SMTDA2016. Valletta, Malta, June 1 - 4, 2016. Book of Abstracts
  1. Christos H Skiadas (ed. lit.)

Éditorial: ISAST: International Society for the Advancement of Science and Technology

ISBN: 978-618-5180-14-0,  978-618-5180-15-7

Année de publication: 2016

Pages: 59-60

Congreso: 4th Stochastic Modeling Techniques and Data Analysis International Conference with 5th Demographics Workshop. SMTDA2016. Valletta, Malta, June 1 - 4, 2016

Type: Communication dans un congrès

GOOGLE SCHOLAR lock_openAccès ouvert editor
Dépôt institutionnel: lock_openAccès ouvert Editor

Résumé

Specifying a regression model requires deciding on which explanatoryvariables it should contain as well as deciding on the functional form thatrelates the dependent variable to the explanatory variables. Variousmodel selection procedures exist. However, there still is no satisfactory answer to the question of which model is “best”. We propose tosupplement model selection procedures with graphical representationsobtained from the techniques of multivariate statistics. The figuresobtained put in evidence the presence of extreme multivariate residuals,displays families of models with similar statistical performance, and canguide model selection for implementation purposes