1
$\begingroup$

I have developed a Cox prognostic model based and internally validated it using bootstrapping. I also evaluated its calibration using Harrell’s $c$ statistic and Somers' $D$.

I have also checked numerous resources, Including Royston P. 2014. Tools for checking calibration of a Cox model in external validation: Approach based on individual event probabilities. Stata Journal 14: 738–755.

Now I want to externally validate my model with another model. I don’t have an individual-level dataset of the model against which I want to validate my model. It is a published Framingham equation in the form:

Predicted risk-at time $t =$

baseline hazard $\times \exp[0.8\ $age $+0.54 $ sex $+0.9\ $ diabetes $+0.5\ $smoking$]$.

I want to use the coefficients in the published Framingham equation (as prognostic index) to validate my model, and subsequently get the $c$ statistics.

According to Tilford JM, Roberson PK, Fiser DH. 1995. Using lfit and lroc to evaluate mortality prediction models. Stata Technical Bulletin 28: 14-18 available here, in logistic regression, I have read that it can be computed by converting this into a matrix:

          mat b = (0.8, 0.54, 0.9, 0.5 )
          mat colnames b = age sex diabetes smoking

Then the post-estimation command lroc can be used to produce the ROC curve and $c$ statistics.

I would really appreciate a suggestion from anyone on how can a $c$ statistic be obtained in a similar scenario for a Cox model.

$\endgroup$
  • $\begingroup$ I've tidied this up but one reference is incomplete and the other is missing. The first line of Stata code is illegal. Requests for Stata code are off-topic here and the emphasis on Stata code would stretch people more accustomed to other software. Still, there seems to be statistical question buried here: can you make it more obvious? $\endgroup$ – Nick Cox Mar 24 '16 at 3:15
  • $\begingroup$ Apologies: The first reference is (Royston P. Tools for checking calibration of a Cox model in external validation: Approach based on individual event probabilities. The Stata Journal 2014; 14(4):738–755. And the second: Tilford JM, Roberson PK, Fiser DH. Using lfit and lroc to evaluate mortality prediction models. Stata Technical Bulletin. Nov 1995. $\endgroup$ – John Mar 24 '16 at 3:52
  • $\begingroup$ Thanks for the references. I have edited the question (which is what you should have done). I've not corrected the Stata error. $\endgroup$ – Nick Cox Mar 24 '16 at 9:23
  • $\begingroup$ I'm really grateful sir @Nick Cox. English is my 2nd (if not the 3rd) that is why I didn't put the question in simplest form. I'll edit the stata syntax. Thanks $\endgroup$ – John Mar 24 '16 at 9:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.