| 讲座简介： | Nonlinear modelling of spatio-temporal data is often a challenge due to irregularly observed locations and location-wide non-stationarity. In this paper we propose a semiparametric family of Dynamic Functional-coefficient Autoregressive Spatio-Temporal (DyFAST) models to address the difficulties to overcome in modelling and analysis. The DyFAST models at least own two significant features. (i) The functional (or varying) coefficient structures that are popular in traditional statistical analysis of i.i.d. and time series data are extended to specify the autoregressive smooth coefficients depending both on a concerned regime and location. The DyFAST models can hence not only characterise the dynamic regime-switching nature but also adapt to the location-wide non-stationarity in real spatio-temporal data. (ii) Two semiparametric smoothing schemes are proposed to model the dynamic neighbouring-time interaction effects with irregular locations incorporated by (spatial) weight matrices. The first scheme that is popular in spatial econometrics supposes that the weight matrix is pre-specified either by experts or by prior information of spatial locations. In practice, the weight matrix may be specified in different ways by data location features. Although model selection for an optimal weight matrix among the candidates is popular, it may suffer from loss of features of different weight matrices. Our second scheme is thus to suggest a weight matrix fusion to let data combine or select the candidates. Accordingly, different semiparametric smoothing procedures are developed for estimation. Both theoretical properties and Monte Carlo simulations are investigated. The empirical application to an EU energy market dataset further demonstrates the usefulness with interesting findings by the DyFAST models. |
