Boolean indexing may be accomplished with both the .iloc and .loc indexers with the caveat that .iloc cannot be passed a Series but the underlying ndarray. Let's take a look at the one-dimensional ndarray underlying the criteria Series:

>>> a = criteria.values
>>> a[:5]
array([False, False, False, False, False], dtype=bool)

>>> len(a), len(criteria)
(4916, 4916)

The array is the same length as the Series, which is the same length as the movie DataFrame. The integer location for the boolean array aligns with the integer location of the DataFrame and the filter happens as expected. These arrays also work with the .loc operator as well but they are a necessity for .iloc.

Steps 6 and 7 show how to filter by columns instead of by rows. The colon, :, is needed to indicate the selection of all the rows. The comma following the colon separates the row and column selections. There is actually a much easier way to select columns with integer data types and that is through the select_dtypes method.

Steps 8 and 9 show a very common and useful way to do boolean indexing on the row and column selections simultaneously. You simply place a comma between the row and column selections. Step 9 uses a list comprehension to loop through all the desired column names to find their integer location with the index method get_loc.