In the presented paper, we introduce a partial functional linear model, where a scalar response variable is explained by a multivariate random variable and a functional random variable, and the relationship between the scalar response and both of the predictors is linear. Besides, the model has autoregressive errors. To estimate the model, we first expand the functional predictor and functional regression parametric on the functional principal component basis, and then estimate the coefficients for multivariate and functional regression parametric by a generalized least squares method. Theoretical properties are presented including the asymptotical normality for the multivariate coefficient and the optimal convergence rate for the functional regression parametric. Simulation studies are used to illustrate these characteristics. The proposed method is also applied on the power forecasting of photovoltaic systems data set.
15 March 2023 in Acta Mathematica Sinica, English Series
Selecting the number of slice is a key step for the implement of the sliced average variance estimation (SAVE) method. To our knowledge, there is no widely accepted method for it in a practical application. And an incorrect number of the slice may leads to an inaccurate conclusion. In traditional multivariate sufficient dimension reduction procedure, it is usually to adopt the fuze approach which combined the kernel operators of SAVE with various numbers of slices to solve this problem. Due to the infinite dimension in functional data, the fuze approach can not be directly applied to functional SAVE (FSAVE). Hence we propose a novel fused approach which based on a weighted kernel operator of FSAVE that named as the approach weight fused FSAVE (WFFSAVE). Simulation studies show that the WFFSAVE method performs better than FSAVE. And we also apply this new method to the Tecator data set.
This paper proposes some novel one-sided omnibus tests for independence between two multivariate stationary time series. These new tests apply the Hilbert-Schmidt independence criterion (HSIC) to test the independence between the innovations of both time series. Under regular conditions, the limiting null distributions of our HSIC-based tests are established. Next, our HSIC-based tests are shown to be consistent. Moreover, a residual bootstrap method is used to obtain the critical values for our HSIC-based tests, and its validity is justified. Compared with the existing cross-correlation-based tests for linear dependence, our tests examine the general (including both linear and non-linear) dependence to give investigators more complete information on the causal relationship between two multivariate time series. The merits of our tests are illustrated by some simulation results and a real example.