An Efficient Approach for Estimating Parameters and Nonparametric Functions in Spatiotemporal Semi-parametric Regression Models
This work investigates statistical inference for both the mean and covariance functions in semiparametric models for complex space–time-dependent data. We propose a new kernel estimator for spatiotemporally correlated data to estimate nonparametric functions, and we embed deep learning methods into the estimation of spatiotemporal covariance. This approach offers considerable flexibility by avoiding restrictive nonseparability and nonstationarity assumptions. We demonstrate that the proposed method improves the bias and efficiency of nonparametric function estimation compared with existing kernel methods such as local linear regression (LLR).
