Janet Netz" <netz@applecon.com> wrtoe,
> I have a series of observations that are censored in two ways: if an
> observation is censored, it may be censored from below or censored from
> above (I observe which type of censoring occurs).
>
> I cannot use the regular tobit command, because it requres that there be the
> same censoring point for each observation. (That is, it calls for a number
> parameter.) For each censored observation, the censor point varies (again,
> I observe what the censor point is).
The command you want is -cnreg-; see [R] tobit in the manual.
-cnreg- allows you to specify observation-by-observation if the outcome is
censored and, if so, whether it is censored from above or below.
You type
. cnreg <yvar> <x variables>, censored(<markervar>)
You must create <markervar> containing 0, -1, and 1. <yvar> and <markervar>
are interpreted jointly,
<yvar> <markervar>
5 0 value y=5 was observed
3 -1 left censoring; y <= 3
2 1 right censoring; y>=2
You could also estimate the model using -intreg-, which has yet a different
syntax:
. intreg y1 y2 <x variables>
In this case, you must create variables y1 and y2 bounding the outcome:
y1 y2
5 5 value y=5 was observed
. 3 left censoring; y <= 3
2 . right censoring; y>=2
-intreg- will also allow interval censoring:
y1 y2
2 4 interval censoring; 2<=y<=4
In any case, -tobit-, -cnreg-, and -intreg- all estimate the same model.
-- Bill
wgould@stata.com
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/