05-06-2014 02:28 PM
I need to model a pooling process similar to a decision tree.. where my data is first separated into products, then whether it's delinquent or not, then a score, this is my current pooling proccess, I guess you can call it a model.
My DV is $Loss. I need to check if adding another field (utilization), does it add any significance to my current model. My DV is $ Loss.
So to achieve this I need to run a regression model and compare the R-squared of the current model to the one where utilization is also added. Any idea of a procedure where I can test this (proc reg didn't allow categorical IV's)? I would like to check my current model, with all it's variables used, and then compare it to the new model where utilization is also added. I have tried a decision tree, but not to friendly with continuous DV's. Thanks.
05-06-2014 03:20 PM
All models convert categorical variables into a coding system via the class statement so you can always do that manually. You should be aware of the method that SAS uses, as there are different methods to create the code and I don't like SAS's default.