Scalable Semiparametric Spatio-temporal Regression for Large Data Analysis

With the rapid advances of data acquisition techniques, spatio-temporal are becoming increasingly abundant in a diverse array disciplines. Here, we develop regression methodology for analyzing large amounts spatially referenced collected over time, motivated by environmental studies utilizing remotely sensed satellite data. In particular, specify semiparametric autoregressive model without usual Gaussian assumption and devise computationally scalable procedure that enables analysis datasets. We estimate parameters maximum pseudolikelihood show computational complexity can be reduced from cubic to linear sample size. Asymptotic properties under suitable regularity conditions further established inform efficient scalable. A simulation study is conducted evaluate finite-sample parameter estimation statistical inference. illustrate our dataset with 2.96 million observations annual land surface temperature, comparison an existing state-of-the-art approach highlights advantages method. Supplementary materials accompanying this paper appear online.