On a Location-Wide Semiparametric Analysis of Spatiotemporal Dynamics of the COVID-19 Daily New Cases in the UK

The COVID-19 pandemic has impacted the way people live worldwide, including the UK. In this paper, we have proposed a location-wide semiparametric spatiotemporal modelling method for analysis of the dynamics of a spatiotemporal daily confirmed number of COVID-19 cases at 367 local authority areas in the UK. Estimation of the spatiotemporal model for the count data taking into account both the nonlinear time trend and the spatial neighboring effect is developed. With the aid of variable selection, it is empirically shown that the proposed model performs well in application to the UK COVID-19 data estimation and prediction. The empirically extracted information from the data provides some new insights into what are the key factors contributing to the confirmed daily number of cases at different locations. It is found that the success of interventions varies depending on location, subject to population, medical resource and role in the national or international transportation network. Our finding also shows that the neighboring effects are significant, and hence, limiting public transportation is always effective to control the spread of the pandemic by reducing contacts. Furthermore, it is empirically noted that the media effects are significant, which may be due to the promotion of self-protection awareness in controlling the spread of the pandemic.