Outlier-robust fit for Generalized Additive Models



This page contains R code implementing a robust fit for Generalized Additive Models, as proposed in: Azadeh, A. and Salibian-Barrera, M. (2011). An outlier-robust fit for Generalized Additive Models with applications to disease outbreak detection. To appear in the Journal of the American Statistical Association (a preliminary version of the manuscript is available here).

  • NEW R package implementing this method is available here (link to CRAN).
  • This package needs the following two packages to be present in your machine:
  • This is joint work with Davor Cubranic

  • IMPORTANT DISCLAIMER:
    This is a preliminarly version of the code.
    Limitations of the current version that I'm aware of include:
    • Leave-one-out cross-validation as it is currently implemented needs a single covariate without ties. This is generally easy to ensure by adding very small noise to the covariate. See the example on the rgam help page.

  • An example script to illustrate the use of the above package is available here.

  • Running the script above generates these plots (click on them to open):

    • Poisson example (1 covariate)

    • Binomial example (1 covariate)

    • Poisson example (2 covariates)

      • True mean surface + points

      • GAM estimated mean surface + points

      • rgam estimated mean surface + points

      • Residuals plot


Please send any comments or feedback to Matias at "matias+stat-ubc-ca".

Last updated Nov 13, 2009.