Stepwise Regression Models

Find a Model Using Variable Selection

The Stepwise personality of the Fit Model platform enables you to fit stepwise regression models, explore all possible models for a set of regressors, and conduct model averaging.

Stepwise regression is an approach to selecting a subset of effects for a regression model. It can be useful in the following situations:

• There is little theory to guide the selection of terms for a model.

• You want to interactively explore which predictors seem to provide a good fit.

• You want to improve a model’s prediction performance by reducing the variance caused by estimating unnecessary terms.

For categorical predictors, you can do the following:

• Choose from among various rules to determine how associated terms enter the model.

• Enforce effect heredity.

Contents

Overview of Stepwise Regression

Example Using Stepwise Regression

The Stepwise Report

Stepwise Platform Options

Stepwise Regression Control Panel

Current Estimates Report

Step History Report

Models with Crossed, Interaction, or Polynomial Terms

Models with Nominal and Ordinal Effects

Construction of Hierarchical Terms

Perform Binary and Ordinal Logistic Stepwise Regression

The All Possible Models Option

The Model Averaging Option

Validation Options in Stepwise Regression

Validation Set with Two or Three Values in Stepwise Regression

K-Fold Cross Validation in Stepwise Regression

Additional Examples of the Stepwise Personality

Example of the Combine Rule

Example of a Model with a Nominal Term

Example of the Restrict Rule for Hierarchical Terms

Example of Logistic Stepwise Regression

Example of the All Possible Models Option

Example of the Model Averaging Option

Want more information? Have questions? Get answers in the JMP User Community (community.jmp.com).