Visualizing Regression Concepts in JMP ®: Picture = (1000*words)
Sep 7, 2017 1:15 PM
Jordan Hiller, JMP Senior Systems Engineer, SAS
This presentation was voted one of the three finalists for Best Invited Paper.
Learning about regression modeling can be challenging. JMP tools such as the Prediction Profiler and the Surface Plot can be particularly useful to help analysts gain insight into modeling concepts visually. I will show brief standalone modules that illustrate and explain the following concepts:
Interaction as warping of the response surface
Use and interpretation of polynomial terms
Thinking about models in 4 or more dimensions
Instability of estimates when regressors are correlated
JMP has always been a great tool for teaching and learning Statistics. Interactive demos and simulations can drive understanding far better than text and static images. There have been many contributions in this area. Most prominent are the teaching resources found within JMP itself, under Help > Sample Data. Also, the JMP Academic team maintains a list of resources here, and at the JMP Academic Community.
This current effort is a group of short interactive modules that can be used to teach tricky concepts in regression modeling, emphasizing visual intuition over mathematical formulas. The modules are aimed at novice analysts who are learning about regression on the job, rather than in the classroom. They use simulated and idealized data, and they are each designed to be completed within 10 minutes.
To use these modules, download the .zip file and unzip them into a folder on your local computer. Each module is launched from its own journal file. The interactive resources are launched from journal hyperlinks. Keep all the journal files and the data tables together in the same folder, to ensure that file references in the hyperlinks remain valid.