In [14]:

    

import numpy
import pandas
import matplotlib.pyplot as plt
import statsmodels.api as sm
import statsmodels.formula.api as smf
import seaborn
import statistics

# bug fix for display formats to avoid run time errors
pandas.set_option('display.float_format', lambda x:'%.2f'%x)

#load the data
data = pandas.read_csv('~/dev/coursera/separatedData.csv')


    

In [15]:

    

# convert to numeric format
data["breastCancer100th"] = pandas.to_numeric(data["breastCancer100th"], errors='coerce')
data["meanSugarPerson"]   = pandas.to_numeric(data["meanSugarPerson"], errors='coerce')
data["meanFoodPerson"]   = pandas.to_numeric(data["meanFoodPerson"], errors='coerce')
data["meanCholesterol"]   = pandas.to_numeric(data["meanCholesterol"], errors='coerce')


    

In [16]:

    

# listwise deletion of missing values
sub1 = data[['breastCancer100th', 'meanSugarPerson', 'meanFoodPerson', 'meanCholesterol']].dropna()

# Create the conditions to a new variable named sugar_consumption that will categorize the meanSugarPerson answers
meanIncidence = statistics.mean(sub1['breastCancer100th'])

def incidence_cancer (row):
    if row['breastCancer100th'] <= meanIncidence : return 0   # Incidence of breast cancer is below the average of the incidence of all countries.
    if row['breastCancer100th'] > meanIncidence  : return 1   # incidence of breast cancer is above the average of the incidence of all countries.

# Add the new variable sugar_consumption to subData
sub1['incidence_cancer'] = sub1.apply (lambda row: incidence_cancer (row),axis=1)

# Create the conditions to a new variable named sugar_consumption that will categorize the meanSugarPerson answers
def sugar_consumption (row):
   if 0 < row['meanSugarPerson'] <= 30 : return 0    # Desirable between 0 and 30 g.
   if 30 < row['meanSugarPerson'] <= 60 : return 1   # Raised between 30 and 60 g.
   if 60 < row['meanSugarPerson'] <= 90 : return 2   # Borderline high between 60 and 90 g.
   if 90 < row['meanSugarPerson'] <= 120 : return 3  # High between 90 and 120 g.
   if row['meanSugarPerson'] > 120 : return 4        # Very high under 120g.

# Add the new variable sugar_consumption to subData
sub1['sugar_consumption'] = sub1.apply (lambda row: sugar_consumption (row),axis=1)


    

In [17]:

    

# Create the conditions to a new variable named food_consumption that will categorize the meanFoodPerson answers
meanFood = statistics.mean(sub1['meanFoodPerson'])

def food_consumption (row):
    if row['meanFoodPerson'] <= meanFood  : return 0   # food consumption below the average of the food consumption of all countries.
    if row['meanFoodPerson'] > meanFood   : return 1   # food consumption above the average of the food consumption of all countries.


    

In [18]:

    

# Add the new variable food_consumption to subData
sub1['food_consumption'] = sub1.apply (lambda row: food_consumption (row),axis=1)

# Create the conditions to a new variable named cholesterol_blood that will categorize the meanCholesterol answers
def cholesterol_blood (row):

   if row['meanCholesterol'] <= 5.2 : return 0         # (0) Desirable below 5.2 mmol/L
   if 5.2 < row['meanCholesterol'] <= 6.2 : return 1   # (1) Borderline high between 5.2 and 6.2 mmol/L
   if row['meanCholesterol'] > 6.2 : return 2          # (2) High above 6.2 mmol/L

# Add the new variable sugar_consumption to subData
sub1['cholesterol_blood'] = sub1.apply (lambda row: cholesterol_blood (row),axis=1)


    

In [19]:

    

# Logistic Regression analysis
lreg1 = smf.logit(formula = 'incidence_cancer ~ sugar_consumption + food_consumption + cholesterol_blood', data = sub1).fit()
print (lreg1.summary())


    

    




Optimization terminated successfully.
         Current function value: 0.273398
         Iterations 8
                           Logit Regression Results                           
==============================================================================
Dep. Variable:       incidence_cancer   No. Observations:                  129
Model:                          Logit   Df Residuals:                      125
Method:                           MLE   Df Model:                            3
Date:                Fri, 07 Oct 2016   Pseudo R-squ.:                  0.5627
Time:                        22:56:34   Log-Likelihood:                -35.268
converged:                       True   LL-Null:                       -80.654
                                        LLR p-value:                 1.496e-19
=====================================================================================
                        coef    std err          z      P>|z|      [95.0% Conf. Int.]
-------------------------------------------------------------------------------------
Intercept            -4.9331      1.049     -4.705      0.000        -6.988    -2.878
sugar_consumption     0.5915      0.317      1.864      0.062        -0.031     1.214
food_consumption      3.0577      0.827      3.696      0.000         1.436     4.679
cholesterol_blood     2.1235      0.650      3.267      0.001         0.849     3.398
=====================================================================================

In [20]:

    

# odd ratios with 95% confidence intervals
params = lreg1.params
conf = lreg1.conf_int()
conf['OR'] = params
conf.columns = ['Lower CI', 'Upper CI', 'OR']
print (numpy.exp(conf))


    

    




                   Lower CI  Upper CI    OR
Intercept              0.00      0.06  0.01
sugar_consumption      0.97      3.37  1.81
food_consumption       4.20    107.69 21.28
cholesterol_blood      2.34     29.90  8.36