Title: Semiparametric ultra-high dimensional model averaging of nonlinear dynamic time series
Authors: Jia Chen - University of York (United Kingdom) [presenting]
Abstract: Two semiparametric model averaging schemes are proposed for nonlinear dynamic time series regression models with a very large number of covariates. The objective is to obtain accurate estimates and forecasts of time series nonparametrically. In the first scheme we use a Kernel Sure Independence Screening technique to screen out insignificant regressors; we then use a semiparametric penalized method of Model Averaging MArginal Regression for the regressors that have survived the screening procedure, to further select regressors that have significant effects on estimating the multivariate regression function and predicting the future values of the response variable. In the second scheme, we impose an approximate factor modelling structure on the ultra-high dimensional exogenous regressors and use the principal component analysis to estimate the latent common factors; we then apply the penalised Model Averaging MArginal Regression method to select significant common factors and lags of the response variable. In each of the two schemes, we construct the optimal combination of the significant marginal regression and auto-regression functions. Asymptotic and numerical studies of the proposed methods are provided.