left censored and survival analysis

[Amparo Nagore]

Please, could you tell me if there is a standard command or a code available in Stata to estimate the survival function taking into account the unobserved heterogeneity, left-censored and right-censored, for instance using Gompertz function? Thank you very much in advance for your help.

[Christophe Kolodziejczyk]

You should be able to deal with left-censoring when you stset your data (se help stset). In this step you tell stata where left-censoring occurs. It is taken care of when you run streg or other st commands. Unobserved heterogeneity can be dealt with to some extent with the frailty option in streg

[Stephen Jenkins]

Amparo: I agree with Christophe's good advice, though I think he meant to say "left truncation" (also known as "delayed entry") rather than "left censoring". The latter feature is very difficult to take account of. Please also note that you should not combine the streg frailty option with left truncation; in fact, by default Stata won't allow you to do so. (The likelihood function for left-truncated data and frailty is very complicated because of the self-selection issues that are involved.) However, see also http://vandenberg.vwl.uni-mannheim.de/2999.0.html?&L=0

[Christophe Kolodziejczyk]

Thanks Stephen for correcting me and giving more insights on the matter. Indeed I meant left truncation, but I must confess I had overlooked the issues related to combining left-trunction combining and unobserved heterogneity. Thanks for the link too