One way to check for more meaningful relationships between data is to add more variables to the model, or in other words, use multiple regression. In a multiple regression, we infer a linear relationship between one response variable and a set of two or more explanatory variables. If your data have three variables (X, Y, and Z), then multiple regression will produce a model of the form:

Z = A + B1X + B2Y

The constant term, A, gives the intercept of the regression equation, and it will have the same units as Z. The slope B1, for example, is the expected difference in Z between two subjects with the same value of Y but a one-unit difference in X. This concept is referred to as controlling for a variable.

Controlling for a variable is a key concept to keep in mind when performing regression diagnostics to both assess the validity of the model and isolate the impact of a single variable. This interactive will cover two types of regression diagnostics: partial residual plots, and partial regression plots.