I estimated two model systems with SAS proc model by using Full information maximum likelihood. Each of them have multiple nonlinear model equations. The two model systems are nested, and the more general one got a lower log likelihood (-25000), while the slightly simpler one has a higher log likelihood (-18000). Since many talks on the internet says the log likelihood should always be higher for the model with more free parameters if the models are nested, I am concerned. So my questions is:
1. Is this normal?
2. Can I still use likelihood ratio test to compare them given this situation?
I am trying to use vuong test to test some non nested model systems. It seems that I need to calculate individual likelihood for each model. My question is:
1. How to calculate individual likelihood in SAS proc model?
Thanks a lot!!!!
Advertise on Talk Stats