Understanding PROC PHREG output

[juliavd]

I am running a Cox-proportional hazards model and using AIC for model selection (I am doing reverse selection, but manually rather than using a program). In the output for AIC, both a "without covariates" and a "with covariates" value are shown. What is the difference between these two values, and which should I be using for model selection?

[SteveDenham]

Without covariates - this is the null model (essentially the information conveyed by the mean alone)

With covariates - this includes any independent variables.  If you are doing some sort of model selection, you should notice a change in the "with covariates" value as the number of independent variables included changes.

SteveDenham