Tree-structured model diagnostics for linear regression

    Authors

    X. G. Su; C. L. Tsai;M. Wang

    Comments

    Authors: contact us about adding a copy of your work at STARS@ucf.edu

    Abstract

    This paper studies model diagnostics for linear regression models. We propose two tree-based procedures to check the adequacy of linear functional form and the appropriateness of homoscedasticity, respectively. The proposed tree methods not only facilitate a natural assessment of the linear model, but also automatically provide clues for amending deficiencies. We explore and illustrate their uses via both Monte Carlo studies and real data examples.

    Journal Title

    Machine Learning

    Volume

    74

    Issue/Number

    2

    Publication Date

    1-1-2009

    Document Type

    Article

    First Page

    111

    Last Page

    131

    WOS Identifier

    WOS:000263062800001

    ISSN

    0885-6125

    Share

    COinS