I am new to spss, but have experience with other packages such as SAS and Statistica.
I am working on a decision tree that has a categorical target variable with 8 different values.
I have converted all my independent variables to numerics - I am only using 7 independent variables.
My problem is that not all 8 values are coming through in the training and test scored data sets. When I build a model with all 8 target values in it, normally only 5 or 6 are predicted in the training and test data sets - the other 3 have zero predictions against them. When I tried a smaller subset model with only 5 of the 8 values represented in the target only 4 were predicted in the training and test data sets. When I reduced it again to a smaller subset containing only 4 of the 8 values all 4 were predicted somewhere in the training and test sets.
The thing is, different values are dropping out depending on the data set I use eg I have a value called Traditional Savers that is predicted in the training and test sets when i use the full input file containing all 8 values and also when i use the subset containing only 4 of the 8, but it doesnt appear in the subset using 5 of the 8.
I have tried transforming the independent variables in many different ways.
Is there something I am missing? Can SPSS not cope with decision trees with 8 different values in a categorical target? Can I force all 8 values to be represented in the predicted output?