2
$\begingroup$

I am using the interval regression in my dissertation (intreg command in Stata), but I can't find any publications describing this particular model. Should I just briefly describe the model in the paper, or there are some references, and I just can't find them?

$\endgroup$
5
  • $\begingroup$ Not directly answering your question, but what type of data do you have? I have written an R-package, icenReg, for survival regression models that I am biased toward recommending. It does not fit a normally distributed linear model (but can fit a log-normal with either proportional odds or hazards model), as stata's package does, so this may or may not be more appropriate for your data, depending on the data type. $\endgroup$ – Cliff AB Jul 14 '15 at 18:55
  • $\begingroup$ My data consists only of points and right-censored observations (a; +Inf) $\endgroup$ – Evgenii Nikitin Jul 15 '15 at 15:39
  • $\begingroup$ These are actually times, so all observations are positive $\endgroup$ – Evgenii Nikitin Jul 15 '15 at 15:54
  • 1
    $\begingroup$ I may be misunderstanding you, but if all the observations are either right censored or actually observed, I would suggest using standard survival regression models (either accelerated failure times or Cox-PH are the most common). Interval censoring is more general than just having right censoring, so using a model that accounts for interval censoring is not invalid. However, most people are more familiar with the right censoring only case, so it should be less difficult to explain. You may have other reasons to use it that I'm overlooking though. $\endgroup$ – Cliff AB Jul 15 '15 at 16:32
  • $\begingroup$ You're actually absolutely right, I realised that I can use survival analysis only after I replied to your initial comment $\endgroup$ – Evgenii Nikitin Jul 16 '15 at 0:48
3
$\begingroup$

The stata help page provides some references. See the bottom of

http://www.stata.com/manuals13/rintreg.pdf

In general, I would recommend briefly describing the methods (or maybe in more detail, given that it is your dissertation rather than a journal article) but also provide references to the original papers describing the methods.

$\endgroup$
1
  • $\begingroup$ Yeah, I chcked references from this page, and if I'm not wrong none of them thoroughly describe the model. I guess I might as well just cite this page itself $\endgroup$ – Evgenii Nikitin Jul 15 '15 at 15:44

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.