Removing the Guesswork for Handling Collinearity in Your Models - Mastering JMP

Video was recorded in March 2025 using JMP 18 and JMP Pro 18.

Do you work with complicated processes where possible unaccounted-for variables might be impacting your predictions? Do you sometimes struggle to identify the validity of factors impacting the results of complicated interactions? Are collinear factors thwarting the efficiency of your process improvement projects?

In this Mastering JMP session we  use machine learning tools for a case study for determining the ideal conditions for cutting precise machine parts for the aviation industry.  See how to:

• Examine the data and uncover possible collinear relationships graphically.
• Explore simple linear regression models and determine why they are not sufficient.
• Use JMP Pro machine learning tools to build and evaluate a model that accurately identifies and handles collinear effects.

Questions answered by @Peter_Polito and @Bill_Worley at the live webinar:

Q: Can PCA and Neural Nets be performed in JMP or do they require JMP Pro?

A: Both are in regular JMP. Neural Nets in JMP offers a single activation function. However,  Neural Nets in JMP Pro also gives you multiple  activation functions, boosting, and model validation.

Q: Does VIF require having repeat measurements in the data?

A: No. And repeat measures are useful instead for understanding variation in models.

Q: Can we use RSM on Principal Components?

A: Yes.

Q: You mentioned Random Forest. Is that an analysis available  in JMP, and if so, can we specify the criterion (i.e. Gini, entropy, etc.)?

A: We call it Bootstrap Forest. It is available in JMP Pro.

Q: Will you please review VIF?

A: Variance Inflation Factor (VIF) is available for each term in the model when the distribution is Normal. High VIF values indicate a collinearity issue among the terms in the model.  To see VIF,  R-click in the parameter estimates table and select Columns > VIF or set it in Preferences for Fit Model options Fit Least Squares, Response Fit and Fit Mixed.  See video below for more on VIF.