Hi,
thanks for your answers Martin and Carlo!
I am a little confused now, do people agree that robust tobit estimation does not account for heteroskedasticity? But what is it then good for?
Best
Phil
-------- Original-Nachricht --------
> Datum: Wed, 4 Mar 2009 11:18:54 -0000
> Von: "Carlo Fezzi" <[email protected]>
> An: [email protected]
> Betreff: st: RE: AW: mfx with tobit
> Phil,
>
> I would not use intreg to fit a Tobit on heteroskedastic data, the
> estimates
> would be biased.
>
> See the related post:
>
> http://www.stata.com/statalist/archive/2006-12/msg00093.html
>
>
> You might be better off using CLAD or writing your own likelihood
> function,
> on this see:
>
> http://www.stata.com/statalist/archive/2007-12/msg00600.html
>
>
> Cheers,
>
> Carlo
>
>
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of Martin Weiss
> Sent: 04 March 2009 11:04
> To: [email protected]
> Subject: st: AW: mfx with tobit
>
>
> <>
>
> Phil,
>
> -mfx- uses the conventions of -predict- after estimation commands, so you
> can obtain information about the options for -mfx- after -tobit- from
>
>
> *************
> help tobit postestimation
> *************
>
> Look for "Syntax for predict".
>
>
>
>
>
> HTH
> Martin
>
>
> -----Ursprüngliche Nachricht-----
> Von: [email protected]
> [mailto:[email protected]] Im Auftrag von
> [email protected]
> Gesendet: Mittwoch, 4. März 2009 11:46
> An: [email protected]
> Betreff: st: mfx with tobit
>
> Hello,
>
> I am running a regression model where the observations are censored from
> the
> left (at zero). Therefore, I use a Tobit model. Due to heteroskedasticity
> present in my data, I use the intreg command to obtain robust standard
> errors. However, I have two questions in this regard.
>
> First, the coefficients obtained from Tobit estimation cannot be directly
> interpreted. I found out that it is possible to use the mfx command to
> obtain the marginal effects and that different options exist to compute
> these effects; e.g. mfx, predict(e(0,1000)) or mfx predict(ystar(0,1000)).
> However, these commands lead to quite different results and I am not able
> to
> figure out which command is preferable in which situation?
>
> Second, I have another dataset where the observations also pile up at
> zero,
> however, in this dataset the dependent variable does not only take on
> positive values, but sometimes also negative values. I am not sure which
> estimation technique is preferable in this case?
>
> I would appreciate any help,
>
> best regards
>
> Phil
>
> --
> Psssst! Schon vom neuen GMX MultiMessenger gehört? Der kann`s mit allen:
> http://www.gmx.net/de/go/multimessenger01
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
--
Psssst! Schon vom neuen GMX MultiMessenger gehört? Der kann`s mit allen: http://www.gmx.net/de/go/multimessenger01
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
