Overview of Variables and Interaction Terms

Variables

Category Variables

Category variables are numeric or nonnumeric variables with discrete levels. The levels of a category variable are considered unordered by SAS Visual Statistics. Examples of category variables include drink size (small, medium, or large), number of cylinders in an engine (2, 4, 6, or 8), or whether a customer has made a purchase (yes or no).
You can create a category variable from a response variable by right-clicking the variable, and selecting Category. In this case, each distinct value of the measure variable is turned into a level for the category variable.
Category variables can be used as response variables for classification models, classification effect variables, decision tree predictors, filter variables, and group by variables.
Note: To ensure proper performance and valid modeling results, the maximum number of distinct levels allowed for a category variable is limited based on the model type and variable role.

Measure Variables

Measure variables are continuous numeric variables that can assume an infinite number of possible values between two numbers. Even though some numeric variables are not continuous, such as count variables, these variables can be treated as continuous values for the purpose of modeling. Examples of measure variables include the temperature of a drink, engine displacement amount, or a customer’s total purchase amount.
Summary statistics and a histogram for each measure variable are obtained by right-clicking the variable in the Data pane, and selecting Properties. Use the Name drop-down menu to specify the variable that you want to view.
Measure variables can be used as response variables for continuous models, continuous effect variables, decision tree predictors, offset variables, frequency variables, weight variables, and filter variables.

Interaction Terms

Two variables, A and B, interact if the effect of one variable on the model changes as the other variable changes. That is, the effects of variables A and B are not additive in the model.
SAS Visual Statistics enables you to create interactions between two or more input variables, including squared interactions. A squared interaction is the interaction of a variable with itself. You cannot create squared interactions for category variables.
For an example where interaction terms might be useful, consider a situation where you are modeling the fuel mileage (MPG) for several cars. Two of your input variables are engine displacement in liters and engine size (number of cylinders). You expect that as either value increases, fuel mileage will suffer. However, if you suspect that the effects on fuel mileage that are attributable to engine displacement are not constant across engine size, then you should consider creating the interaction term between those variables.
SAS Visual Statistics is not limited to creating just two-way interactions. You can create n-way interactions that include an arbitrary number of variables, but not more than the number of available input variables.
The number of distinct levels for an interaction term is the product of the number of levels for each variable in the term. Measure variables are treated as if they contain one level. The number of levels in an interaction term counts against the maximum number of distinct levels allowed in regression models.