Using ADF test to check stationarity

ADFTest.py

# Loading the packages
import pandas as pd
import numpy as np
import statsmodels.tsa.stattools as sm
import matplotlib.pyplot as plt
plt.style.use('fivethirtyeight')

# Loading the dataset:
data = pd.read_csv('../AirPassengers.csv')
data = data.rename(columns = {'#Passengers':'Passengers'})
data = data.set_index('Month')

# Plotting the data (Figure A):
ax = data.plot(linewidth = 0.8) 
ax.set_xlabel('Year')
ax.set_ylabel('Number of Passengers')
ax.set_title('Monthly Passengers')

# Applying ADF Test:
nonst_test = sm.adfuller(data)

# Printing the results (Table under Figure A):
output = pd.Series(nonst_test[0:4], index=['t-statistic','p-value','lags-used','no-of-observations'])
for key,value in nonst_test[4].items():
    output['Critical Value (%s)'%key] = value
  
print(output)

# Apply transformation to make data stationary:
log_data = np.log(data) # Taking the log
ma_data = log_data.rolling(window=12).mean() # Taking moving average
log_minus_ma_data = log_data - ma_data
log_minus_ma_data.dropna(inplace=True)

# Plotting the data (Figure B):
ax = log_minus_ma_data.plot(linewidth = 0.8)
ax.set_xlabel('Year')
ax.set_ylabel('Number of Passengers')
ax.set_title('Monthly Passengers')

# Applying ADF Test after applying transformations:
st_test = sm.adfuller(log_minus_ma_data)

# Printing the results (Table under Figure B):
output = pd.Series(st_test[0:4], index=['t-statistic','p-value','lags-used','no-of-observations'])
for key,value in st_test[4].items():
    output['Critical Value (%s)'%key] = value
  
print(output)