Title: Partially linear time series models with time-varying coefficients
Authors: Lionel Truquet - ENSAI (France) [presenting]
Abstract: Partially linear models have been widely studied in Econometrics and Statistics. Their semiparametric nature is general enough for capturing complex nonlinearities and meanwhile it avoids the curse of dimensionality. In linear regressions with time-varying coefficients, partially linear models appear naturally when testing time-constancy of some of the parameters, or for the semiparametric inference in such models. We will discuss how to conduct inference in this case, using in particular Robinsons type estimates. In a second part, we will discuss an extension of partially linear models using time-varying coefficients as an interesting approach for avoiding the curse of dimensionality in nonparametric, time-varying regression models.