In [1]:

    

import numpy as np
import pickle
from scipy.stats import pearsonr
import sys

%matplotlib inline
import matplotlib.pyplot as plt

%load_ext autoreload
%autoreload 2

sys.path.append('../')
import scripts.outliers as outliers


    

In [5]:

    

DATA_FILE = '../data/lda_data_8.pickle'
METADATA_FILE = '../data/metadata.csv'
dataset, ddf, w_dict = outliers.load_data(DATA_FILE, METADATA_FILE)
X_list, Y, Yaudio = dataset
X = np.concatenate(X_list, axis=1)


    

    




Antigua and Barbuda
Australia
Cuba
Fiji
French Polynesia
Grenada
Iceland
Jamaica
Japan
Kiribati
Malta
New Zealand
Philippines
Puerto Rico
Republic of Serbia
Saint Lucia
Samoa
Solomon Islands
South Korea
The Bahamas
Trinidad and Tobago

In [6]:

    

df_global, threshold, MD = outliers.get_outliers_df(X, Y, chi2thr=0.999)
df_global.head()


    

    
Out[6]:






  

    

      

      
Country
      
Outliers
      
N_Country
      
N_Outliers
    
  
  

    

      
0
      
Canada
      
0.050000
      
100
      
5
    
    

      
1
      
Lithuania
      
0.000000
      
47
      
0
    
    

      
2
      
Cambodia
      
0.210526
      
19
      
4
    
    

      
3
      
Ethiopia
      
0.257143
      
35
      
9
    
    

      
4
      
Swaziland
      
0.163265
      
98
      
16
    
  

In [10]:

    

corr, pval = pearsonr(df_global['Outliers'], df_global['N_Country'])
print 'correlation', corr
print 'p-value', pval

plt.scatter(df_global['Outliers'], df_global['N_Country'])
plt.xlabel('Outliers');
plt.ylabel('N');


    

    




correlation -0.0102335874359
p-value 0.905523601988






    

In [ ]: