Programming the statistical procedures from SAS

How to build a deep learning model in SAS Enterprise Miner

Occasional Contributor
Posts: 9

How to build a deep learning model in SAS Enterprise Miner

I have been working in SAS for about 3-4 months, and during that time, SAS has been very helpful in constructing multiple Z-Tests, T-Tests and Chi-Squares for me on my data. However, now that I have finished those, I'm looking for a conclusive way to use all of the data from the results of the tests I've run. My mentor has recommended that I build a Proc Glimmix model, so that I can take multiple different variables and compare them to predict an outcome. I need this model because I need a receiver operator curve, as well as a good way to cross validate subsets. My overall goal is to do a logistic regression, fabricate a logistic probability of being in any category by doing binary categorization, and have an overall generalized linear model. I am aware that if I build my model with too many variables I run the risk of overfitting it, but I also need enough variables to take care of variance. 


I am not really certain how I should approach building my Proc Glimmix Model because neither I nor my mentor have any experience in that area. I would greatly appreciate if someone could supply me with examples of Proc Glimmix Model code, any step by step manual that SAS may have (I have yet to find one that seems relevent to my research), and any advice that I may have to correct my code.


Any help is appreciated!

Super User
Posts: 20,765

Re: How to build a deep learning model in SAS Enterprise Miner

The doc has 19 examples, is there one that's close to your situation?


Rather than focus on the procedure I think you should be focusing on the methods you're interested in and then finding the appropriate procedure. From your description, PROC LOGISTIC sounds appropriate. 


Otherwise, two other resources are:

UCLA tutorials under Statistics section


And is full of user written papers on almost any procedure. 


Ask a Question
Discussion stats
  • 1 reply
  • 2 in conversation