Variable and boundary selection for functional data via multiclass logistic regression modeling

Access this Article

Search this Article

Abstract

L1 penalties such as the lasso provide solutions with some coefficients to be exactly zeros, which lead to variable selection in regression settings. They also can select variables which affect the classification by being applied to the logistic regression model. We focus on the form of L1 penalties in logistic regression models for functional data, especially in the case for classifying the functions into three or more groups. We provide penalties that appropriately select variables in the functional multinomial regression modeling. Simulation and real data analysis show that we should select the form of the penalty in accordance with the purpose of the analysis.

Journal

  • Computational Statistics & Data Analysis

    Computational Statistics & Data Analysis (78), 176-185, 2014-10

    Elsevier

Codes

  • NII Article ID (NAID)
    120005527740
  • NII NACSIS-CAT ID (NCID)
    AA10002391
  • Text Lang
    ENG
  • Article Type
    journal article
  • ISSN
    0167-9473
  • Data Source
    IR 
Page Top