It is late in my part of the world: The probability of being censored
is ofcourse dependent on the unobserved value of the censored variable
itself...
so you are left with with my latter two remarks.
Maarten
--- In statalist@yahoogroups.com, "maartenbuis" <maartenbuis@y...> wrote:
> ---Chris Rohlfs <car@u...> wrote:
> > Does anyone out there know of a way to deal with censored covariates
> > in Stata ? (i.e., something like cnreg, but where it's a regressor
> > that's censored)
>
> Chris:
> If you can assume that the probability of being censored is
> independent from the (obviously unobserved) value of the censored
> variable itself, than you can use standard missing data techniques
> like multiple imputation using the -mvis- command or multiple hotdeck
> imputation using the -hotdeck- command. If you can not, than you can
> ignore the censored cases (with loss of efficiency, but without bias)
> if the probability of being censored is independent of dependent
> variable (the y, or explained variable), since regression will produce
> unbiased estimates as long as the probability of being censored is
> independent from the dependent variable. To my best knowledge you are
> out of luck (with Stata or any other statistical package) if the
> probability of being censored is dependent on both the censored
> variable itself and the dependent variable.
>
> Hope this helps,
> Maarten
>
>
> *
> * 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/
*
* 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/