Attributes with Category Sets

Categorical attributes can’t have units, but they can have category sets assigned to them. A category set defines the valid values that such an attribute can have. The bar chart at right, for example, shows that the valid values of Race are “Asian”, “Black”, “Latino”, and “White”. The bar chart leaves a place for “Latino” even though there are no people of that race in the collection. The bar chart also lumps all the values that are not in the category set into a single “<invalid>” group. Using category sets can be very helpful, and sometimes absolutely necessary, but you can go a long way in Fathom without them.

When an attribute has a category set assigned to it, Fathom will treat it as a categorical attribute whether or not it has any non-numeric values. The attribute for year in the U.S. census microdata, for example, is automatically assigned a category set. That way, year can be used as a splitting attribute. Plotting year versus income will give you, not a scatter plot, but a split dot plot.

Assigning a category set to an attribute also ensures that the data, when plotted on a graph or put in a summary table, will be arranged by the order defined in the category set, rather than alphabetically.