It is well documented that the presence of outliers and/or extreme observations can have a strong impact on smoothing spline. This has motivated the development of robust smoothing procedures. In particular, some studies have focused on the robust selection of smoothing parameter proposing extensions of the generalized cross-validation method. In this work we consider an alternative for accommodation of outliers in spline smoothing. Our proposal is based in to consider the penalty introduced in smoothing splines as a random effect. We use a nested EM algorithm to perform the parameter estimation under distributions with tails heavier than normal. Numerical example and a simulation study illustrate the techniques. We expect that this approach allows us to choose the smoothing parameter automatically and can be seen as an alternative to cross-validation.

Keywords: Penalized likelihood; Smoothing splines; Outliers; Nested EM algorithm

Biography: Felipe Osorio is Assistant Professor at the Department of Statistics of the University of Valparaiso, Chile. His research interests include: Sensitivity analysis and influence diagnostics and Mixed-effect models.