We consider a high-dimensional model in which variables are observed over time and space. The model consists of a spatio-temporal regression containing a time lag and a spatial lag of the dependent variable. Unlike classical spatial autoregressive models, we do not rely on a predetermined spatial interaction matrix, but infer all spatial interactions from the data. Assuming sparsity, we estimate the spatial and temporal dependence fully data-driven by penalizing a set of Yule-Walker equations. This regularization can be left unstructured, but we also propose customized shrinkage procedures when observations originate from spatial grids (e.g. satellite images). Finite sample error bounds are derived and estimation consistency is established in an asymptotic framework wherein the sample size and the number of spatial units diverge jointly. Exogenous variables can be included as well. A simulation exercise shows strong finite sample performance compared to competing procedures. As an empirical application, we model satellite measured NO2 concentrations in London. Our approach delivers forecast improvements over a competitive benchmark and we discover evidence for strong spatial interactions.
翻译:我们考虑的是一种高维模型,在时间和空间中观测变量。该模型包括时空回归,包含一个时滞和依附变量的空间时滞。与传统的空间自动递减模型不同,我们并不依赖一个预先确定的空间互动矩阵,而是从数据中推断出所有空间互动。假设空间宽度,我们通过惩罚一套Yule-Walker方程式来充分估算空间和时间依赖性数据。这种规范化可以保持不结构化,但是当观测来自空间网格(例如卫星图象)时,我们也提议定制缩缩缩程序。精确样本误差是推断出来的,并在一个以样本大小和空间单位数目相异的单一框架中确定估算一致性。外源变量也可以被包含在内。模拟演练表明,与竞争程序相比,样本的性能是相当有限的。作为经验应用,我们模拟卫星测量了伦敦的NO2浓度。我们的方法预测了在竞争性基准上作出的改进,并发现了强大的空间互动证据。