05-02-2013 12:06 AM
I have two models that I want to compare using the Pseudo-R Squared measure: 100 x (Null Deviance - Residual Deviance)/Null Deviance .
I'm unsure which deviance measure to use, the "Deviance" or "Scaled Deviance" once I have run each model (Null Model : Intercept Only, and Model of Interest : Desired Covariates). The documentation mentions the scaled deviance being twice the difference of two different log-likelihoods - I'm guessing that this is for use in a Likelihood Ratio test, and probably not the deviance value I want to use in the above formula.
Thanks in advance for any help.