Predict COVID-19 Cases by County-Level Socio-Economics Indicators

• Activity Number: 75 - SPEED: Data Challenge and SPAAC
• Type: Topic-Contributed
• Date/Time: Monday, August 9, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Section on Statistical Computing
• Author(s): Yue Yu* and Shiying Wang
• Companies: University of Michigan and Georgetown University
• Keywords: COVID-19 Cases Prediction; Public Policy Evaluation; American Community Survey ; Quadratic Spline ; Variable Selection ; Time Series Analysis

Abstract

This study evaluates the relationship between county-level socio-economics indicators, which are from the 2019 American Community Survey (ACS) 1-year Estimates, and COVID-19 new cases in the United States, to evaluate the effectiveness of past COVID-related public policies, to further predict the COVID-19 new cases, and to provide policy implement suggestions at county-level. 7 degree-of-freedom quadratic spline models were used to capture COVID-19 new cases features from March 1, 2020 to December 31, 2020. Variable selection methods such as LASSO and XGBoost were used to pick highly important variables from over 500 variables in ACS raw data. Meanwhile, time series analysis models, like Autoregressive Moving Average (ARMA) and Exponential Smoothing (ES), were used to bring further insights for prediction.