Multivariate Adaptive Regression Splines (MARSplines) Startup Panel - Options Tab
Select the Options tab of the Multivariate Adaptive Regression Splines (MARSplines) Startup Panel to access options to specify the maximum complexity (number of basis functions) and stop-criterion for the iterative model building procedure implemented in MARSplines. Refer to the Introductory Overview and Technical Notes for more details.
| Element Name | Description |
|---|---|
| Model specifications | This group box contains the following options: |
| Maximum number of basis functions | In this field, specify the maximum number of basis functions to be added to the MARSplines model before pruning begins; for details, refer to the Introductory Overview and Technical Notes. |
| Degree of interactions | In this field, specify the maximum degree of interactions between the independent (predictor) variables to be included in the model; for example, enter 1 to include main-effects only, enter 2 to consider all two-way interactions between predictors, etc. |
| Penalty | In this field, specify the value of the penalty for adding a basis function to the model. The penalty value is used in computations of the least-squares goodness-of-fit measure that guides the MARSplines algorithm; for details, see Technical Notes: The Multivariate Adaptive Regression Splines (MARSplines) Model. |
| Threshold | This option prevents overfitting; the larger the threshold, the fewer basis functions will be added to the regression model. |
| Apply pruning | Select this check box to apply pruning after fitting the requested number of basis functions to avoid overfitting, i.e. to limit model complexity. It is generally recommended to apply pruning in order to achieve a lower generalized cross validation error (i.e., predictive validity for new cases). |
Copyright © 2021. Cloud Software Group, Inc. All Rights Reserved.