08-16-2016 01:47 AM
I am new here. The proc lifereg used to produce a model fit statistic (loglikelihood) without any of the covariates along the usual one with covariates. This is useful to compute a likelihood ratio or a pseudo r-squares. It seems I cannot find this now with newer version of SAS/ETS. Is there a solution for this? I cannot run the model with no covariates as I do not know what missing observations to exclude to have a match between the two runs (with and without covariates). Thank you in advance for any help.
08-16-2016 08:38 PM
So I checked a couple of different version of the documentation (your welcome) and I can't see a difference in the output. That doesn't mean there isn't one, just nothing related to Log Likelihood that I can find.
Can you post sample code (and ideally sample data) that you think should generate log likelihood and a reference to what version of SAS would produce this value.
You can use the data from the documentation as sample data if your data is confidential.