In [267]:

    

import readline
import rpy2.robjects
import code.independence_test as it
import code.additive_noise as an
import numpy as np
import numpy.random as rn
import code.hsic as hs
import pandas as pd


    

In [268]:

    

n=250
cause = rn.uniform(0,10,n)
indep = rn.uniform(0,10,n)

n_Y = np.random.uniform(-1,1,n)
n_L = np.random.uniform(-1,1,n)
n_C = np.random.uniform(-1,1,n)


    

In [269]:

    

Y = np.log(cause) + n_Y 
confounded = np.log(cause) + n_C
leaked = Y**2 + n_L


    

In [270]:

    

X = np.stack([cause, indep, confounded, leaked],axis=1)
X = pd.DataFrame(X, columns=['cause', 'indep', 'confounded', 'leaked'])


    

In [271]:

    

Y[0:10]


    

    
Out[271]:





array([-1.2285647 ,  0.94596792,  2.13165018,  1.49199779, -0.66854162,
        2.56994782,  1.15495431,  2.63818878,  0.33733907,  2.5795691 ])

In [272]:

    

X.head()


    

    
Out[272]:






  

    

      

      
cause
      
indep
      
confounded
      
leaked
    
  
  

    

      
0
      
0.219043
      
1.567766
      
-1.499400
      
2.130497
    
    

      
1
      
3.411820
      
0.944898
      
0.737476
      
0.450120
    
    

      
2
      
9.513069
      
3.038724
      
3.132322
      
4.678155
    
    

      
3
      
2.386470
      
1.142832
      
1.189391
      
3.099337
    
    

      
4
      
0.822754
      
0.301655
      
-0.995328
      
0.406533
    
  

In [273]:

    

e = an.ANM_algorithm_with_test(X,Y, 'py', test='mi-g-sh')


    

    




Number of line searches 40
Number of line searches 40
Number of line searches 40
Number of line searches 40
Number of line searches 40
Number of line searches 40
['cause'] 2.02022978633e-10
['indep'] 1.05073264159e-34
['leaked'] 9.37036151922e-39
['cause', 'indep'] 3.16233015318e-10
['cause', 'leaked'] 7.39449057411e-16
['indep', 'leaked'] 6.63048328346e-32
['cause', 'indep', 'leaked'] 3.7867038201e-15
Number of line searches 40
Number of line searches 40
['cause'] 2.22298081076e-06
['indep'] 6.22865689541e-21
['confounded'] 5.54478650181e-22
['cause', 'indep'] 2.11216349294e-06
['cause', 'confounded'] 6.86366186717e-12
['indep', 'confounded'] 2.80908201314e-18
['cause', 'indep', 'confounded'] 2.41792446901e-11

In [274]:

    

e


    

    
Out[274]:






  

    

      

      
Direction
      
Rank
      
P_value Independence X and Y
    
  
  

    

      
indep
      
Independent
      
0.00342176
      
0.148439
    
    

      
leaked
      
Leakage
      
0.00230293
      
1.6922e-79
    
    

      
confounded
      
Leakage
      
0.000538001
      
2.65347e-18
    
    

      
cause
      
No Leakage
      
-0.0132915
      
4.82163e-36
    
  

In [141]:

    

e.ix['confounded'].Direction


    

    
Out[141]:





"Column: confounded is independent of target given ['cause']"

In [ ]: