A Macro-Driven Approach for Systematically Testing Variables Against a Base Regression Model
Jennifer Alessandro, Bryan Harmon · 2012
Have you ever had a base regression model and wanted to test what happens when you add a variable such as television advertising to that base model? What if you have 10 different television variables, each with four different possible retention rates? You may even want to try some of the advertising variables in combination with one another. In the end, only one or a small subset of these variables should be added to the base model. All of them need to be tested, but testing them in the model all at once will cause collinearity issues. This paper will provide a macro you can use to test a large set of variables quickly and efficiently. The output will provide a summary showing how each candidate variable worked in the base model and which combinations of candidate variables worked well together. It will also go into detail on how to easily generalize the macro, so it can be used for many projects.