Logistic Regression Model Properties
The following properties are available for the
logistic regression model:
Name
enables you to specify
the name for this model.
Significance level
specifies the
significance level that is required in order for variables to be considered for the model. This property
is available only when
Use variable
selection is selected.
Link function
The following link
functions are available:
-
Logit (default)
specifies the inverse of the cumulative logistic distribution function.
-
Probit specifies
the inverse of the cumulative standard normal distribution function.
Convergence
-
Override function convergence enables
you to manually specify the function convergence value.
-
Value specifies
the function convergence value when Override function
convergence is selected. When you specify a larger value,
the model will converge sooner. This reduces the amount of time spent
training the model, but it can create a suboptimal model.
-
-
Value specifies the gradient convergence value when Override gradient
convergence is selected. When you specify a larger value,
the model will converge sooner. This reduces the amount of time spent
training the model, but it can create a suboptimal model.
-
Maximum iterations specifies the maximum number of
iterations performed during model training. If you specify a relatively small value, you reduce
the amount of time spent training the model, but it can create a suboptimal model.
Note: When you specify a gradient
convergence or function convergence criterion, it is possible for
the model to converge based on an internal convergence criterion before
your specified criterion is reached. The reason for convergence is
on the Convergence tab of the details table.
Assessment
-
Use default number
of bins specifies whether you want to use the default number of
bins or to set your own value. By default, measure variables are grouped into 20 bins.
-
Number specifies the number of bins to use when the
Use default number
of bins property is not selected. You must specify an
integer value between 5 and 100.
-
Prediction cutoff specifies
the value at which a computed probability is considered an event.
-
Show diagnostic plots
specifies whether the Residual
Plot, Assessment, and Influence
Plot windows appear in the model pane.
Copyright © SAS Institute Inc. All rights reserved.