1. Read the college dataset with the institution name as the index:

>>> college = pd.read_csv('data/college.csv', index_col='INSTNM')
>>> college.dtypes
CITY                   object
STABBR                 object
HBCU                  float64
MENONLY               float64
                       ...   
PCTFLOAN              float64
UG25ABV               float64
MD_EARN_WNE_P10        object
GRAD_DEBT_MDN_SUPP     object
Length: 26, dtype: object

2. All the other columns besides CITY and STABBR appear to be numeric. Examining the data types from the preceding step reveals unexpectedly that the MD_EARN_WNE_P10 and GRAD_DEBT_MDN_SUPP columns are of type object and not numeric. To help get a better idea of what kind of values are in these columns, let's examine their first value:

>>> college.MD_EARN_WNE_P10.iloc[0]
'30300'

>>> college.GRAD_DEBT_MDN_SUPP.iloc[0]
'33888'

3. These values are strings but we would like them to be numeric. This means that there are likely to be non-numeric characters that appear elsewhere in the Series. One way to check for this is to sort these columns in descending order and examine the first few rows:

>>> college.MD_EARN_WNE_P10.sort_values(ascending=False).head()
INSTNM
Sharon Regional Health System School of Nursing    PrivacySuppressed
Northcoast Medical Training Academy                PrivacySuppressed
Success Schools                                    PrivacySuppressed
Louisiana Culinary Institute                       PrivacySuppressed
Bais Medrash Toras Chesed                          PrivacySuppressed
Name: MD_EARN_WNE_P10, dtype: object

4. The culprit appears to be that some schools have privacy concerns about these two columns of data. To force these columns to be numeric, use the pandas function to_numeric:

>>> cols = ['MD_EARN_WNE_P10', 'GRAD_DEBT_MDN_SUPP']
>>> for col in cols:
        college[col] = pd.to_numeric(college[col], errors='coerce')

>>> college.dtypes.loc[cols]
MD_EARN_WNE_P10       float64
GRAD_DEBT_MDN_SUPP    float64
dtype: object

5. Use the select_dtypes method to filter for only numeric columns. This will exclude STABBR and CITY columns, where a maximum value doesn't make sense with this problem:

>>> college_n = college.select_dtypes(include=[np.number])
>>> college_n.head()

6. By utilizing the data dictionary, there are several columns that have only binary (0/1) values that will not provide useful information. To programmatically find these columns, we can create boolean Series and find all the columns that have two unique values with the nunique method:

>>> criteria = college_n.nunique() == 2
>>> criteria.head()
HBCU          True
MENONLY       True
WOMENONLY     True
RELAFFIL      True
SATVRMID     False
dtype: bool

7. Pass this boolean Series to the indexing operator of the columns index object and create a list of the binary columns:

>>> binary_cols = college_n.columns[criteria].tolist()
>>> binary_cols
['HBCU', 'MENONLY', 'WOMENONLY', 'RELAFFIL', 'DISTANCEONLY', 'CURROPER']

8. Remove the binary columns with the drop method:

>>> college_n2 = college_n.drop(labels=binary_cols, axis='columns')
>>> college_n2.head()

9. Use the idxmax method to find the index label of the maximum value for each column:

>>> max_cols = college_n2.idxmax()
>>> max_cols
SATVRMID                      California Institute of Technology
SATMTMID                      California Institute of Technology
UGDS                               University of Phoenix-Arizona
UGDS_WHITE                Mr Leon's School of Hair Design-Moscow
                                         ...                    
PCTFLOAN                                  ABC Beauty College Inc
UG25ABV                           Dongguk University-Los Angeles
MD_EARN_WNE_P10                     Medical College of Wisconsin
GRAD_DEBT_MDN_SUPP    Southwest University of Visual Arts-Tucson
Length: 18, dtype: object

10. Call the unique method on the max_cols Series. This returns an ndarray of the unique column names:

>>> unique_max_cols = max_cols.unique()
>>> unique_max_cols[:5]
array(['California Institute of Technology',
       'University of Phoenix-Arizona',
       "Mr Leon's School of Hair Design-Moscow",
       'Velvatex College of Beauty Culture',
       'Thunderbird School of Global Management'], dtype=object)

11. Use the values of max_cols to select only the rows that have schools with a maximum value and then use the style attribute to highlight these values:

>>> college_n2.loc[unique_max_cols].style.highlight_max()