Classifying Texas counties using ARIMA Models on COVID-19 daily confirmed cases: the impact of political affiliation and face covering orders

Jiang, H.; Pulins, B.; Thiele, A.

The aim of this paper is to investigate whether the 254 Texas counties in the United States can be grouped in a meaningful way according to the characteristics of the ARIMA or seasonal ARIMA models fitting the logarithm of daily confirmed cases of the Coronavirus Disease 2019 (COVID-19) for 254 counties in Texas of the United States. We analyze clusters of the model's non-seasonal parameters $(p,d,q)$, distinguishing between county-level political affiliations and face covering orders, and also consider county-level population and poverty rate. Using data from March 4, 2020 to March 15, 2021, we find that 223 of the total 254 counties are clustered into 23 model parameters $(p,d,q)$, while the number of cases in the remaining 31 counties could not be successfully fitted to ARIMA models. We also find the impact of the county-level infection rate and the county-level poverty rate on clusters of counties with different political affiliations and face covering orders. Further, we find that the infection rate and the poverty rate had a significant high positive correlation, and Democrat-leaning counties, which tend to have large populations, had a higher correlation coefficient between infection rate and poverty rate. We also observe a significant high positive correlation between the infection rate and the number of cumulative cases in Republican counties that had not imposed a face covering order.