Regression Model to Predict Tire Performance: Paired Design Experiment

1 Kudo

JMP Regression Model to Predict Tire Performance: Demo and Tutorial

Leighton Spadone, DAAS Inc.

Tire Technology and Statistical Consulting

See Attachments: Updated 21Sep2014 R1

JMP Conference 2014 Paper Abstract

Overview:

This is Interactive Regression Modeling Tutorial.

Uses Paired Design to predict Tire Performance from Lab wheel test.

This is “Real World” Problem Solving with Cost Savings Benefits.

“Aha!” Solution is to Model and Predict the Percent-Change in the value!

Create and Compare 2 Regression models for their usefulness.

Check Model Data Quality by analysis of Outliers and Marginal Plots.

Presentation:

Define Giant Off-the-Road Tire performance measurement “TMPH”.

Show the Modeling Objective, Paired Design Methodology, and Cost Benefits.

Describe Data Table and the benefits of the Paired Design: Power and Size.

Data Table Exploration using Fix X by Y and Tabulate.

Describe modeling prediction limits due to restricted sample size.

Describe “Aha” Solution to predict Percent-Change in TMPH to improve the model’s usefulness by avoiding extrapolation.

Create and Compare 2 different models usefulness.

Search for Outliers using Control Chart and Mahalanobis Distance.

Handling Outliers discussion.

Compare model with and without Outliers.

Create Marginal Plots for model.

JMP Interactive Tools:

Fit X by Y

Tabulate

Standard Least Squares

Control Chart

Mahalanobis Distance

Data Filter

Marginal Plots