15  Types of Attributes

In CODAP, most attributes are in one of several categories. The most important of these are:

15.1 How CODAP gets it wrong

For the most part, CODAP interprets your data values the way you expect. But there are exceptions.

Numeric variables that are really categorical. In the original Titanic data set, the survived attribute is 0 for “died” and 1 for “lived.” That’s just the way the data were entered; it was easier to enter ones and zeros than typing in “died” or “lived” so many times.

Or imagine a dataset about students with an attribute called class that holds their graduation year, so that a value of 2025 means, “Class of 2025.” This is a numeric variable, but really its underlying purpose is categorical: every case gets put into one of (usually four) categories.

Categorical variables that are really numeric. Sometimes, data are entered with text and numbers, but you really care only about the numbers. For example, in some Census data, if you ask for number of children, you get values such as, “0 children present” and “1 child present,” but then “2”, “3”, and so forth:

Census data. Notice the values of the Number_of_children attribute.

Often, any text that’s not a number will fool CODAP into thinking it’s categorical. In the illustration, all it takes is the dollar signs to mess things up. The graph is not only categorical, it’s in the order of the table instead of numerical order!

Dollar signs can fool CODAP into thinking these numbers are categories.

15.2 How to fix it

There are three ways to address this problem.

Attribute properties box

One is to change the type in the attribute properties box. Click on the column heading to get the menu, then choose Edit Attribute Properties…. The box appears; choose a type in the type menu.

You can also change the name, specify units, give a description, tell how many decimal places you want, and restrict editing!

Using the type menu…

…to change the money attribute to numeric.

In the graph

The axis labels on a graph are also menus. You can use the menus to change which attribute is displayed, but there’s more: down at the bottom, you can change the way the graph displays the attribute with the commands Treat as Categorical or Treat as Numeric. As an example, here is some Titanic data where 1 means “survived” and 0 means “died”:

Left: survived is treated as numeric

Right: survived treated as categorical

Making a new attribute

Continuing with the Titanic example, suppose we wanted to display the values as survived and perished instead of as 1 and 0.

Then we could use the technique described in the section about recoding categorical attributes.

To recap,

  1. Drag survived to the left to make groups of each value.
  2. Make a new attribute next to survived (maybe call it didTheyLive or something suitable).
  3. Type survived next to the 1 and perished next to the 0.
  4. Substitute the new attribute for survived on a graph.

It will look like this:

Titanic survival, with a new attribute giving values we understand.

15.3 In the cracks: Ordered categoricals, discrete numerics

Some sets of values, by their nature, share some properties of both categorical and numeric attributes.

For example, ages in Census data are always whole numbers. Even though people can be 20.75 years old, in the data you’re 20 until you turn 21. One consequence of this is that, although will generally group data (by dragging left in the table) with categorical attributes, and never with decimals, there are situations when whole-number numerics are perfectly good for grouping. That’s exactly what we did in one of our first lessons with age and height.

And some categorical values, e.g., {frosh, soph, junior, senior} have a natural order. When CODAP has no clue about the order, it alphabetizes them. But you can change that by re-ordering the values on a graph. You might do that when exploring education in Census data.

  1. Despite our growing suspicion that there is really no such thing as a categorical attribute, they can be really convenient.↩︎