TSMethods.py

# Time Series
import numpy as np
import matplotlib.pyplot as plt

def create_series(df, xcol, datecol):
    # Create a dataframe with the features and the date time as the index
    features_considered = [xcol]
    features = df[features_considered]
    features.index = df[datecol]
    # features.head()
    # features.plot(subplots=True)
    return features


# X is the series to test
# log_x asks whether to log X prior to testing or not
def stationarity_test(X, log_x = "Y", return_p = False, print_res = True):
    
    # If X isn't logged, we need to log it for better results
    if log_x == "Y":
        X = np.log(X[X>0])
    
    # Once we have the series as needed we can do the ADF test
    from statsmodels.tsa.stattools import adfuller
    dickey_fuller = adfuller(X)
    
    if print_res:
    # If ADF statistic is < our 1% critical value (sig level) we can conclude it's not a fluke (ie low P val / reject H(0))
        print('ADF Stat is: {}.'.format(dickey_fuller[0]))
        # A lower p val means we can reject the H(0) that our data is NOT stationary
        print('P Val is: {}.'.format(dickey_fuller[1]))
        print('Critical Values (Significance Levels): ')
        for key,val in dickey_fuller[4].items():
            print(key,":",round(val,3))
            
    if return_p:
        return dickey_fuller[1]
    
# Differencing the data    
def difference(X):
    diff = X.diff()
    plt.plot(diff)
    plt.show()
    return diff