In [ ]:

    

# note, there are placeholders throughout the code, that you must properly insert
# for entire script to work


    

In [1]:

    

%matplotlib inline
import matplotlib.pyplot as plt
import pandas as pd
import statsmodels.api as sm
import pylab as pl
import numpy as np


    

In [19]:

    

df_raw = pd.read_csv("../assets/admissions_test.csv")
df = df_raw.dropna() # drop any rows, with a null value for any column
print df.dtypes # look carefully at what columns are present in this dataFrame
df.count()
# df.columns
# pd.crosstab(df['admit'],df['prestige'])


    

    




admit         int64
prestige    float64
dtype: object






    
Out[19]:





admit       192
prestige    192
dtype: int64

In [8]:

    

df_no_admit = df[df['admit'] == 0]
rows_no_admit, cols_no_admit = df_no_admit.shape
print rows_no_admit


    

    




138

In [9]:

    

df_admit = df[df['admit'] == 1]
rows_admit, cols_admit = df_admit.shape
print rows_admit


    

    




54

In [11]:

    

print (15./rows_admit)*100


    

    




27.7777777778

In [10]:

    

print (12./rows_no_admit)*100


    

    




8.69565217391

In [15]:

    

df_no_admit.hist()
df_admit.hist()


    

    
Out[15]:





array([[<matplotlib.axes._subplots.AxesSubplot object at 0x118946d10>,
        <matplotlib.axes._subplots.AxesSubplot object at 0x1189cb850>]], dtype=object)






    













    

In [17]:

    

#create a frequency table using cross tab from pd, what columns do we want to group by here?
# group_column_1 = ?
# group_column_2 = ? 

pd.crosstab(df['admit'],df['prestige'])


    

    
Out[17]:






  

    

      
prestige
      
1.0
      
2.0
      
3.0
      
4.0
    
    

      
admit
      

      

      

      

    
  
  

    

      
0
      
12
      
44
      
50
      
32
    
    

      
1
      
15
      
25
      
10
      
4
    
  

In [23]:

    

df.prestige.plot(kind = 'density')


    

    
Out[23]:





<matplotlib.axes._subplots.AxesSubplot at 0x119055710>






    

In [24]:

    

df.prestige.mean()


    

    
Out[24]:





2.546875

In [25]:

    

df.prestige.std()


    

    
Out[25]:





0.9533026774138411

In [27]:

    

df_no_admit.prestige.plot(kind = 'density')


    

    
Out[27]:





<matplotlib.axes._subplots.AxesSubplot at 0x119507750>






    

In [28]:

    

df_admit.prestige.plot(kind = 'density')


    

    
Out[28]:





<matplotlib.axes._subplots.AxesSubplot at 0x119592c50>






    

In [36]:

    

# create dummy variables using pd.create_dummies
# When your variables are categorical, you should create a dummy varaible. Subjectivity is key.
# When you can quantify variables, ordinal variables, you might not need to have dummy vars.
# dummy_column = ?
# prefix_for_dummy_col = ?

dummy_ranks = pd.get_dummies(df['prestige'], prefix='prestige')
print dummy_ranks.head(4)

# prestige = [3,3,1,4]


    

    




   prestige_1.0  prestige_2.0  prestige_3.0  prestige_4.0
0             0             0             1             0
1             0             0             1             0
2             1             0             0             0
3             0             0             0             1






    
Out[36]:





192

In [20]:

    

cols_to_keep = [] # insert column(s) that you want to keep in your dataFrame here before the join
df_ready = df[cols_to_keep].join(dummy_ranks)


    

In [ ]:

    

# 3.1 Use the cross tab above to calculate the odds of 
# being admitted to grad school if you attended a #1 ranked college
# group_by_column_1 = ?
# group_by_column_2 = ?

pd.crosstab(df_ready['group_by_column_1'],df_ready['group_by_column_2'])


    

In [24]:

    

# if you are using Logit() from statsmodel.linear_model, you must specify y-intercept (constant)
df_ready['intercept'] = 1


    

In [ ]:

    

# make a new dataFrame with only your predictors, 
#in our case just dummy variables, minus baseline prestige_4
predictor_cols = [] # insert final columns as string delimtied by comma that you want to keep in here
df_predictors = df_ready[predictor_cols]
df_predictors.head(2)


    

In [ ]:

    

# use sm.Logit() from statsmodels.api to build logit model
# rmr class column is what we are trying to 'predict'
# your_class_column = ?

logit = sm.Logit(df_ready['your_class_column'], df_predictors)
results = logit.fit()


    

In [ ]:

    

# get model coefficients, Confidence Intervals for Predictors etc, via the .summary method
# on a Logit model
results.summary()


    

In [ ]:

    

model_coef = results.params # to get model coefficients
print model_coef


    

In [ ]:

    

# to calculate odds of coefficients after model, rmr to use np.exp(coefficients)
print np.exp(model_coef)


    

In [60]:

    

# creating fake data for predictions
prestige_test_set = np.random.randint(1,4,50) # give me 50 numbes between 0 and 4 inclusive in list
df_test = pd.DataFrame({'prestige':prestige_test_set, 'intercept': 1}) # make dataFrame


    

In [ ]:

    

df_test.describe()


    

In [ ]:

    

# dummy_column = ? 
# dummy_column_prefix = ? 

dummy_ranks_test = pd.get_dummies(df_test['dummy_column'], prefix='dummy_column_prefix') # create dummies for test df
dummy_ranks_test.columns


    

In [ ]:

    

test_cols_to_keep = [] #insert string of columns, comma delimited, that you want to keep as predictors 
df_for_pred = df_test.join(dummy_ranks_test)[test_cols_to_keep] #join then select only pred cols
df_for_pred.head(2)


    

In [ ]:

    

df_for_pred['admit_pred'] = results.predict(df_for_pred) # new col with predicted admit based on model

print df_for_pred.tail()
len(df_for_pred)


    

In [ ]:

    

# note how a probability is always returned in admit_pred (Logistic Regression)