B0336
Title: Partially linear single-index models: A robust approach
Authors: Ana Maria Bianco - Universidad de Buenos Aires (Argentina) [presenting]
Maria Florencia Statti - Universidad de Buenos Aires (Argentina)
Abstract: When using fully nonparametric models, practitioners often face the curse of dimensionality. In this context, dimension reduction becomes a relevant issue. Partially linear single-index models are a good strategy to reduce dimension and capture nonlinear trends simultaneously. These models are a reasonable trade-off between the fully parametric and fully non-parametric approaches. We propose a robust two-stage estimation procedure of the parametric and nonparametric components of the model when the scale parameter is unknown. We study the consistency of the estimators and derive the asymptotic distribution of the linear and single index parameters. A simulation study is performed, and an application to a real dataset is illustrated. We also explore the finite sample properties of a Wald-type test to check hypotheses that involved the linear parameter.
