I am not an expert on Tobit, or even a user of it. But
there is surely folklore on this problem, which sounds as
if it could arise frequently.
You're nudging 0 to 1, in a context where expenditure is presumably
zero or positive. And presumably 1 is (substantially?) less than
the smallest positive number observed.
The nudge is still a fudge. What you choose could vary drastically
on a log scale, which is crucial here. In other contexts, some people
use (half smallest non-zero number observed). In yet other contexts,
people loop over a range of fudges and see what difference
it makes.
Nick
n.j.cox@durham.ac.uk
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu]On Behalf Of fatma bircan
> Sent: 21 April 2004 13:41
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: st: RE: ln1=0 for missing values
>
>
> For example in a Tobit model if we make a log transformation of the
> expenditures (dep. var.) some values of the dep. var. will have
> missing values since they were originally equal to zero. But assuming
> that the zero exp. =1 and then taling the log of expenditures, we can
> estimate the tobit model. Is this OK for having extra grouns?
>
>
> -------------------
> >Yes, there are. Unless you have extra grounds
> >for thinking that missing really means 1,
> >this practice surely cannot be recommended
> >in general.
> >
> >Nick
> >n.j.cox@durham.ac.uk
> >
> >fatma bircan
> >
> >> Is there any problems with assuming that the dependent variable
> takes
> >> the value of 1 when it is missing to make the ln(dep. var.) equal
> to
> >> zero.
> >
*
* 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/