See how to interactively and iteratively build and choose useful models.
Presenter: Bill Worley
The presenter uses a molded chairs case study where the goal is a 95% yield. He uses Fit Model, Standard Least Squares and Stepwise Regression to build models. To avoid introducing possible irrelevant interactions, he starts by determining if main effects are sufficient. He compares actual to predicted outcomes and uses Effects Summary to add terms and eliminate terms not contributing to the model. He shows how to use reported statistics, the Prediction Profiler and Residual by Predicted Plots to compare models. Q&A from the original live webcast is included at the end.
The presenter starts by fitting one categorical response. He demonstrates how to use the Receiver Operating Characteristics Curve, Confusion Matrix and Prediction Profiler Desirability to evaluate and select the best model. Q&A from the original live webcast is included at the end.
Back to Top