SAVE Subcommand (RBF command)

The SAVE subcommand writes optional temporary variables to the active dataset.

PREDVAL (varname varname…). Predicted value or category. If a dependent variable has scale measurement level, then this is the predicted value. If a dependent variable is categorical, then this is the predicted category.

  • Specify one or more unique, valid variable names. There should be as many variable names specified as there are dependent variables, and the names should be listed in the order of the dependent variables on the command line. If you do not specify enough variable names, then default names are used for any remaining variables. If you specify too many variable names, then any remaining names are ignored.
  • If there is only one dependent variable, then the default variable name is RBF_PredictedValue. If there are multiple dependent variables, then the default variable names are RBF_PredictedValue_1, RBF_PredictedValue_2, etc., corresponding to the order of the dependent variables on the command line.

PSEUDOPROB (rootname:n rootname…). Predicted pseudo-probability. If a dependent variable is categorical, then this keyword saves the predicted pseudo-probabilities of the first n categories of that dependent variable.

  • Specify one or more unique, valid variable names. There should be as many variable names specified as there are categorical dependent variables, and the names should be listed in the order of the categorical dependent variables on the command line. The specified names are treated as rootnames. Suffixes are added to each rootname to get a group of variable names corresponding to the categories for a given dependent variable. If you do not specify enough variable names, then default names are used for any remaining categorical dependent variables. If you specify too many variable names, then any remaining names are ignored.
  • A colon and a positive integer giving the number of probabilities to save for a dependent variable can follow the rootname.
  • If there is only one dependent variable, then the default rootname is RBF_PseudoProbability. If there are multiple dependent variables, then the default rootnames are RBF_PseudoProbability_1, RBF_PseudoProbability_2, etc., corresponding to the order of the categorical dependent variables on the command line and taking into account the position of any scale dependent variables.
  • The default n is 25.
  • This keyword is ignored for scale dependent variables.

Probabilities and Pseudo-Probabilities

Predicted pseudo-probabilities cannot be interpreted as probabilities because the RBF procedure uses the sum-of-squares error and identity activation function for the output layer. The SAVE subcommand saves these predicted pseudo-probabilities even if any are less than 0 or greater than 1, or the sum for a given dependent variable is not 1.

The ROC, cumulative gains, and lift charts (see /PLOT ROC, GAIN, and LIFT, respectively) are created based on pseudo-probabilities. In the event that any of the pseudo-probabilities are less than 0 or greater than 1, or the sum for a given variable is not 1, they are first rescaled to be between 0 and 1 and to sum to 1. The SAVE subcommand saves the original pseudo-probabilities, but the charts are based on rescaled pseudo-probabilities.

Pseudo-probabilities are rescaled by dividing by their sum. For example, if a case has predicted pseudo-probabilities of 0.50, 0.60, and 0.40 for a three-category dependent variable, then each pseudo-probability is divided by the sum 1.50 to get 0.33, 0.40, and 0.27.

If any of the pseudo-probabilities are negative, then the absolute value of the lowest is added to all pseudo-probabilities before the above rescaling. For example, if the pseudo-probabilities are –0.30, 0.50, and 1.30, then first add 0.30 to each value to get 0.00, 0.80, and 1.60. Next, divide each new value by the sum 2.40 to get 0.00, 0.33, and 0.67.