Other Features and Properties

Discover some other features, properties, and caveats of handling categorical data.


Having covered the essential operations and methods around categorical data, let's wrap up this chapter by going over some other noteworthy features and properties.

Unioning categories

To combine multiple categorical variables with different categories, we must first create a common set of categories for them. We can do so with the union_categoricals() function, which generates a union of the categories being combined. It works with Series, Categorical, and CategoricalIndex, and the output of the union operation is a Categorical object.

When we refer to data types, there are two concepts at play:

  • DataFrame columns (or Series objects) can have different dtypes. When dealing with a categorical variable, we can encode it with dtype='category'. For such situations, we’ll use the term dtype.

  • For categorical variables that are already encoded with dtype='category', the category elements can have different data types too. For example, the Education column contains integer values indicating the different education levels, whereas the Ethnicity column contains string values indicating the various ethnicity groups. For such situations, we’ll use the term data type.

An important thing to note is that for union_categoricals() to work, all the categories must have the same data type. For example, we can successfully combine two category-encoded Series objects with categories of the same integer data type.

Get hands-on with 1200+ tech skills courses.