Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: zero-inflated Tobit models


From   Katie Farrin <kfarrin@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: zero-inflated Tobit models
Date   Mon, 10 Jun 2013 06:51:41 -0400

Nick,

I think Austin's advice is wonderful if, in fact, profits are my main
variable of interest.  I will certainly look into it once I have
access to Stata 13 - but that might not be until August.  I'm unsure
why you think running a Tobit - even if it may be incorrectly
specified - would be a bad idea in the first stage of 2SLS.  As is,
the first stage is OLS, which is often an incorrect specification.  I
was under the impression that if I use bootsrapped standard errors I
could use a variety of first-stage model specifications in 2SLS to
improve second-stage efficiency.  I am simply trying to account for
the fact that most of my outcome variables in the first stage take on
values of zero.  I attempted the use of two separate instruments - for
the zero-one choice to operate a business and for the continuous
profits that would result from the choice to operate - but have a weak
instruments problem there.

Best,
Katie

On Mon, Jun 10, 2013 at 6:28 AM, Nick Cox <njcoxstata@gmail.com> wrote:
> I think you are downplaying very good advice from Austin here. If you
> have any negative values, a tobit with censoring at zero sounds wrong
> in principle. Trying to complicate an incorrect model does not seem
> like a good strategy.
>
> Nick
> njcoxstata@gmail.com
>
>
> On 10 June 2013 11:18, Katie Farrin <kfarrin@gmail.com> wrote:
>> Thanks, Austin,
>>
>> I inquired about zero-inflated models because the regression I'm
>> performing is a first-stage regression in an IV model.  I am
>> instrumenting for business profits due to nonrandom selection into
>> owning a business.  I merely wanted a better fit to improve the
>> significance of my second-stage results (I'm bootstrapping my errors),
>> and thus am not sure I need GMM or SEM at this point in my research.
>> While the data isn't technically censored at zero, I have only 4 (out
>> of 1900) observations where profits are to the left of zero.  I have a
>> lot of zero observations (about 1500),
>>
>> Again, does anyone know of a zero-inflated Tobit command (or option
>> within a command) in Stata?
>>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/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/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index