Abstract:
|
This project aims to study patterns in national responses to the pandemic and their effect on controlling the outbreak through a nonparametric density and regression technique and various clustering tools. A novel nonparametric density technique was utilized to consistently partition seven different countries’ COVID-19 cumulative growth curves into waves and phases, and each phase was modeled using linear regression. Every country’s model variables (the slope, r-square value, duration, and model type of each phase) were then connected to real-life factors, such as cluster outbreaks and government regulations. Multivariate correlation was conducted to uncover the relationships between model variables, and variable clustering showcased which model variable would be a good predictor for the infection situation in a certain phase. Finally, based on the multivariate correlation and variable clustering results, the most important variables were used in hierarchical clustering to identify and explain the most similar and most different countries.
|