Quick tab
Select the Quick tab of the Lasso Regression results dialog box to access the options described here.
NOTE: The graph outputs available from the Quick tab are presented with a top x scale indicating the degrees of freedom of each model.
Element Name | Description |
---|---|
All Coefficients | Click this button to produce a spreadsheet of the matrix of coefficients for each step of the coordinate descent along with the lambda values. |
Graph of coefficients | Click this button to produce a graph with plots of all coefficients against log of lambda value at each coordinate descent step. |
Regularization path | Click this button to produce a spreadsheet of stepwise details of coordinate descent. |
Graph of deviance | Click this button to produce a graph with plots of all coefficients against fractional deviance at each coordinate descent step. |
L1 Norm | Click this button to produce a graph with plots of all coefficients against L1 norm at each coordinate descent step. |
Error plot | Click this button to produce a graph with plot of average error against log of lambda values. The graph will contain the indications of the minimum error and error within one standard deviation of the minimum. |
Model results | This group of results applies to a specific step in the coordinate descent. |
Lambda | Specify an approximate lambda value from the coordinate descent steps for which you want to get specifics. |
Coefficients | Click this button to produce a spreadsheet that contains the parameter estimates for the model specified by the lambda value. |
Predictions | Click this button to produce a spreadsheet that contains the predicted values for the model specified by lambda value. Observed are actual dependent variable values from initial dataset. |
Save bundle | Click this button to save the categorical and continuous predictors (as bundles to the input spreadsheet) included in the model specified by the
lambda value.
Options / C / W / By Group. See Common Options. |
Copyright © 2021. Cloud Software Group, Inc. All Rights Reserved.