In [1]:

    

import pandas_datareader.data as pdr
import statsmodels.api as sm
import numpy as np

end = '2016/9/30'
n225 = pdr.DataReader("NIKKEI225", 'fred', '1949/5/16', end).dropna()
lnn225 = np.log(n225.dropna())
lnn225.columns=['Close']
y = lnn225
x = range(len(lnn225))
x = sm.add_constant(x)
model = sm.OLS(y, x)
results = model.fit()
results.summary()


    

    




C:\Users\tono\Anaconda3\lib\site-packages\statsmodels\compat\pandas.py:56: FutureWarning: The pandas.core.datetools module is deprecated and will be removed in a future version. Please use the pandas.tseries module instead.
  from pandas.core import datetools






    
Out[1]:





OLS Regression Results

  
Dep. Variable:
          
Close
      
  R-squared:         
 
   0.756
 


  
Model:
                   
OLS
       
  Adj. R-squared:    
 
   0.756
 


  
Method:
             
Least Squares
  
  F-statistic:       
 
5.196e+04


  
Date:
             
Tue, 24 Oct 2017
 
  Prob (F-statistic):
  
  0.00
  


  
Time:
                 
20:29:52
     
  Log-Likelihood:    
 
 -18601.
 


  
No. Observations:
      
 16769
      
  AIC:               
 
3.721e+04


  
Df Residuals:
          
 16767
      
  BIC:               
 
3.722e+04


  
Df Model:
              
     1
      
                     
     
 
    


  
Covariance Type:
      
nonrobust
    
                     
     
 
    




    
       
coef
     
std err
      
t
      
P>|t|
  
[0.025
    
0.975]
  


  
const
 
    6.1967
 
    0.011
 
  546.859
 
 0.000
 
    6.174
 
    6.219


  
x1
    
    0.0003
 
 1.17e-06
 
  227.943
 
 0.000
 
    0.000
 
    0.000




  
Omnibus:
       
560.110
 
  Durbin-Watson:     
 
   0.000


  
Prob(Omnibus):
 
 0.000
  
  Jarque-Bera (JB):  
 
 284.479


  
Skew:
          
-0.119
  
  Prob(JB):          
 
1.68e-62


  
Kurtosis:
      
 2.408
  
  Cond. No.          
 
1.94e+04

In [3]:

    

%matplotlib inline
import matplotlib.pyplot as plt
plt.plot(y, label='Close', color='darkgray')
results.fittedvalues.plot(label='prediction', style='--')
plt.ylabel('log(n225 index)')
plt.legend(loc='upper left')


    

    
Out[3]:





<matplotlib.legend.Legend at 0x1fb8fb7fa58>






    

In [4]:

    

results.resid.plot(color='seagreen')
plt.ylabel('residual')


    

    
Out[4]:





<matplotlib.text.Text at 0x1fb8fe2da90>






    

In [5]:

    

results.resid.hist(bins=100, color='lightgray')
plt.xlabel('residual')
plt.ylabel('frequency')


    

    
Out[5]:





<matplotlib.text.Text at 0x1fb8fb532b0>






    

In [6]:

    

# 戦後復興期
y = lnn225.ix[:'1954/11/30'].dropna()
x=range(len(y))
x=sm.add_constant(x)
model=sm.OLS(y,x)
results=model.fit()
results.summary()


    

    
Out[6]:





OLS Regression Results

  
Dep. Variable:
          
Close
      
  R-squared:         
 
   0.762


  
Model:
                   
OLS
       
  Adj. R-squared:    
 
   0.761


  
Method:
             
Least Squares
  
  F-statistic:       
 
   4438.


  
Date:
             
Tue, 24 Oct 2017
 
  Prob (F-statistic):
  
  0.00
 


  
Time:
                 
20:43:11
     
  Log-Likelihood:    
 
 -65.389


  
No. Observations:
      
  1391
      
  AIC:               
 
   134.8


  
Df Residuals:
          
  1389
      
  BIC:               
 
   145.3


  
Df Model:
              
     1
      
                     
     
 
   


  
Covariance Type:
      
nonrobust
    
                     
     
 
   




    
       
coef
     
std err
      
t
      
P>|t|
  
[0.025
    
0.975]
  


  
const
 
    4.5236
 
    0.014
 
  332.550
 
 0.000
 
    4.497
 
    4.550


  
x1
    
    0.0011
 
 1.69e-05
 
   66.618
 
 0.000
 
    0.001
 
    0.001




  
Omnibus:
       
160.558
 
  Durbin-Watson:     
 
   0.003


  
Prob(Omnibus):
 
 0.000
  
  Jarque-Bera (JB):  
 
 121.287


  
Skew:
          
 0.624
  
  Prob(JB):          
 
4.60e-27


  
Kurtosis:
      
 2.268
  
  Cond. No.          
 
1.60e+03

In [8]:

    

plt.plot(y, label='Close', color='darkgray')
results.fittedvalues.plot(label='prediction', style='--')
plt.ylabel('log(n225 index)')
plt.legend(loc='upper left')


    

    
Out[8]:





<matplotlib.legend.Legend at 0x1fb90192748>






    

In [ ]: