04-07-2015 02:39 PM
I was wondering if there would be a reason for a Negative Binomial distribution to take hours and hours to process (have yet to let it finish), when applying a poisson distribution to the same model, processes in about 5 mins when using Proc Glimmix?
04-07-2015 03:49 PM
In addition, I reduce the size of the data to 58,522 records. I reran the model using the poisson and then attempted to run the Negative Binomial. This time SAS 9.4 kicked me an insufficient memory error, but will allow me to run the Poisson model again after giving me the error. I have a 16 GB machine, but maybe I'm missing something. I still don't understand why the NB would consume so much more memory than a Poisson????