Now before we cheat and look at our answer key, let's see how well this solution does at predicting data it hasn't seen. To do this, I write the following fairly large test:
def final_model_cross_validation_test():
df = pandas.read_csv('./generated_data.csv')
df['predicted_dependent_var'] = 25.6266 \
+ 2.7083*df['ind_var_a'] \
- 1.5527*df['ind_var_b'] \
- 0.3917*df['ind_var_c'] \
- 0.2006*df['ind_var_e'] \
+ 5.6450*df['ind_var_b'] * df['ind_var_c']
df['diff'] = (df['dependent_var'] - df['predicted_dependent_var']).abs()
print df['diff']
print '==========='
cv_df = pandas.read_csv('./generated_data_cv.csv')
cv_df['predicted_dependent_var'] = 25.6266 \
+ 2.7083*cv_df['ind_var_a'] \
- 1.5527*cv_df['ind_var_b'] \
...