CMStatistics 2015: Start Registration
View Submission - CMStatistics
B0711
Title: Using P-splines for variable selection in functional regression Authors:  M Carmen Aguilera-Morillo - Universidad Carlos III de Madrid (Spain) [presenting]
Rosa Lillo - Universidad Carlos III de Madrid (Spain)
Juan Romo - Universidad Carlos III de Madrid (Spain)
Abstract: The focus is on the selection of variables in functional linear models (FLM) with functional response and scalar covariates. Specifically, a computationally efficient version of functional LASSO is proposed in terms of the basis representation of the functional response variable, since it is well known that LASSO is a powerful method to select variables and shrink parameters in linear models. If the original data are functions with noise, it is advisable to use P-splines in order to improve the estimation of the linear model and the corresponding forecasts. In addition, it is shown that P-splines can also be a useful tool in variable selection. In order to find the best method to carry out both variable selection and estimation of the parameters in the function-on-scalar regression models, functional LASSO, ordinary least squares and P-spline penalty are combined providing interesting results.