Proc Lifereg parameter estimates

Reply
TFD
N/A
Posts: 1

Proc Lifereg parameter estimates

I am new to parametric models and would appreciate some assistance regarding SAS outputs in LIFEREG. When fitting a parametric model using the DIST = WEIBULL, which parameters in the "...Likelihood Parameter Estimates" table pertain to AFT and which pertain to proportional hazards formulation?

I am specifically looking to calculate the hazard at any time:

                    

                           h(t) = lambda*scale*t^scale-1

Lambda and baseline hazard for the proportional hazards assumption should be e^-1*intercept. However, I've read some documentaton that states that for PH formulation all estimates need to be multiplied by the shape parameter before exponentiation. Does this include the intercept?

Scale is either the "Scale" or "Weibull Scale." Certainly the "Weibull Scale" I'm receiving is not consistent with the very low hazard seen in our sample of patients using KM curves.

The code I'm using is:

proc lifereg;

   model time*censor(0)= /dist = Weibull *no covariates in the model as we are currently interested in modeling without covariates;

run;

Results (more in attachment)

Intercept 4.6458

Scale 1.1118

Weibull Scale 130.5

Weibull Shape 0.8994

Lastly, is there a SAS procedure to plot the survivor function directly using the shape and scale parameters?

Please see attachements for other output.

Thanks

Ask a Question
Discussion stats
  • 0 replies
  • 732 views
  • 0 likes
  • 1 in conversation