Linear Lasso Regression: Options

The Options tab provides options for:

Mode

This selection provides options for specifying one of the following modes:

Fit with specified alpha

When you select this mode, a single model is fitted to the training data by using only one alpha regularization value. This is set by default. If a partition is specified, the single or final model that is fitted is applied to the held-out test data to estimate out-of-sample performance.

Under Plot, plots of observed and/or residuals versus predicted values can be selected.

Under Save, you can specify predicted values and residuals to save.

Trace plot

When you select this mode, three plots for the training data are displayed as a function of alpha for the specified set of alpha values:

A trace plot of regression coefficients.
A plot of R².
A plot of mean squared error (MSE).

Although the partition is honored, no results for held-out test data are provided because no final model results from this mode.

Alpha selection via cross validation

When you select a grid search with cross validation to evaluate models, and choose the best alpha based on the best average, R² over the validation folds. The Number of cross validation folds field can be used to change the default value of five splits or folds for cross validation. If a partition is specified, the single or final model that is fitted is applied to the held-out test data to estimate out-of-sample performance.

Under Display, you can select to show basic information about only the model with the chosen value of alpha (Best), basic information about all models compared (Compare models), or complete information on all splits or folds for all models (Compare models and folds). Best is the default.

Under Plot, plots of mean R² and/or MSE over validation folds can be selected. Plots of observed and/or residuals versus predicted values can also be selected.

Under Save, you can specify predicted values and residuals to save.

Specify individual alphas

When you select the Fit with specified alpha mode, you can specify a single alpha regularization value. When you select the Trace plot or Alpha selection via cross validation mode, you can specify multiple values.

Value(s)

Specify one or more positive alpha regularization values. Multiple values can be specified individually or as ranges. The default is 1.

Specify grid of alpha values

When you select the Trace plot or Alpha selection via cross validation mode, a grid of unique alpha values can be specified from a Start value (value1) to an End value (value2) with the increment of By (value3). If specified, only one valid set of [value1 TO value2 BY value3] is allowed. It must satisfy that 0 ≤ value1 ≤ value2 ≤ 1. In cases where

value1 =
value2

, it is equivalent to specifying a single value1, regardless of value3.

The Alpha metric for ranges of values can be either Linear or Base 10 logarithmic (10 is raised to the power of specified values).

Plots are displayed by using the specified metric for the horizontal X axes of the varying alpha values.

Criteria

Controls analyses.

Include intercept: Includes an intercept in the fitted model(s). Note that the extension procedure does not center or standardize the dependent variable, and the intercept is not penalized during estimation.
Standardize predictors: Standardizes all independent variables.
Number of cross validation folds: The number of splits or folds for cross validation evaluation of models. Must be a positive integer value greater than 1. The default is 5.
Python random state: The value of the random_state setting in Python used when performing cross validation evaluation of models. Allows replication of results that involve pseudo-random numbers. Must be an integer in the range of 0 to 2³²-1. The default is 0.
Time limit (minutes): The number of minutes allowed for the model computations to run. If you specify 0, the timer is turned off. The default value is 5.

Display

Specifies the amount of output to display for the Alpha selection via cross validation mode.

Best: Displays only basic results for the chosen best model. This is the default.
Compare models: Displays basic results for all evaluated models.
Compare models and folds: Displays full verbose results for each split or fold for each evaluated model.

Plot

Specifies plots of observed or residual values versus predicted values, and with cross validation, specification of plots of average mean squared error (MSE) and/or average R² over cross validation folds versus alpha values.

Average cross validation mean squared error (MSE) versus alpha: For the Alpha selection via cross validation mode, displays a line plot of average MSE over cross validation folds versus alpha. For the Trace plot mode, a similar plot is automatically produced based on the complete training data.
Average cross-validation R Square versus alpha: For the Alpha selection via cross validation mode, displays a line plot of average R² over cross validation folds versus alpha. For the Trace plot mode, a similar plot is automatically produced based on the complete training data.
Observed versus Predicted: Displays a scatterplot of observed versus predicted values for the specified or best model.
Residuals versus Predicted: Displays a scatterplot of residuals versus predicted values for the specified or best model.

Save

Specifies variables to save to the active data set.

Predicted values: Save predicted values from the specified or best model to the active data set. You can also specify a Custom variable name.
Residuals: Save residuals from the specified or best model predictions to the active data set. You can also specify a Custom variable name.