Build and Apply QSAR/QSPR Model — Best Subsets Options Pane

Set parameters for determining the best subsets of the X variables to use in a multiple linear regression. The best subsets are determined by a simulated annealing Monte Carlo simulation. To open this pane, select Multiple Linear Regression in the Method option menu in the Build Task, and click the Best Subsets link in the Multiple Linear Regression Settings section.

Best Subsets Options Pane Features

Subset size text box

Set the number of X variables in the subsets.

Return average of N best models option and text box

Select this option to use the average of a specified number of the best models for the best subset, and specify the number of models.

Weight by R^2 option

When averaging models, weight the models by their R2 value. Only available when Use average of N best models is selected.

Simulated Annealing section

Set options for the simulated annealing calculation.

Number of Monte Carlo steps text box

Set the number of Monte Carlo steps in the simulated annealing process used to determine the best subsets.

Initial temperature text box

Set the initial “temperature” for the simulated annealing process as a multiple of the standard deviation in the Y variable.

Final temperature text box

Set the final “temperature” for the simulated annealing process as a multiple of the standard deviation in the Y variable.