Experiments being performed to prove the hypothesis "Covid is a patchwork pandemic in the US" for the Geographic Review.

1. Random forest regression on Arizona data. The regression is followed by a prediction over test data for all the states. A series of R^2 values are stored.
2. Random forest regression on each state over a set of time lags. The regression is followed by a prediction over test data for each instance. R^2 values are tabled.
3. Vector Auto Regression (VAR) over the 'change in deaths' time series. Predictions are made over the test data using the fitted model. R^2 values are tabled.

Note - This experiment is run with and without exogenous data ('change in m50_index').

4. Vector Error Correction Model (VECM) over the 'change in deaths' time series. Predictions are made for <number of prediction steps(k)> using the fitted model (hence the model has to be fitted repeatedly upto nth day,for predicting k days after nth day ). R^2 values are tabled.

Note - This experiment is run with and without exogenous data ('change in m50_index').

5. Forecasting change in m50_index using the VAR modeling as in exp3.
6. Granger causality between the change in mobility and change in deaths over a series of lags for a chosen window.

Steps -

1. Manually copy file from drive and save as 'data/United_States_COVID-19_Cases_and_Deaths_by_State_over_Time.csv'.
2. Manually copy file from drive and save as 'data/Mobility_upto 28-09-2020_DL.xlsx'.