Re: st: tobit interpretation

[David Jones <djones123.listserver@gmail.com>]

Hi Christine,

it is important to remember that with nonlinear models such as tobit,
probit,etc, the coefficient is not directly interpretable. There are
some cases (I'm not sure off the top of my head!) where the sign might
be directly interpretable (such that an increase in your independent
causes a decrease in your dependent). What you are truly interested in
is the marginal effect of a change in the variable. This can be easily
found by using the command margins. For more information about the
syntax, look at the help file.

It is important to note, however, that there are multiple values for
which you might be interested in the marginal effect of. You may look
at the mean of the variable, you may look at a specific value that is
of concern to your research (such as the increase from zero to one -
ie zero participation to participation). It may be helpful to read
about marginal effects here:
http://qed.econ.queensu.ca/faculty/abbott/econ452/452note03_new_slides.pdf

I don't know your level of sophistication with calculus, so that may
not be helpful, but I hope I've started to give you an idea!

Best,
David

On Wed, Jul 31, 2013 at 10:29 PM, Christine E. Boyle
<christine@bluehorizoninsight.com> wrote:
> Hello,
>
> I am estimating the effect of village education level (portion of villagers
> who attended middle school) on share of village investment by the villagers
> using a tobit estimator.  Dependant and independent variables are both
> between 0 and 1.
>
> Once I set the   ul (1) and the ll(0), negative coefficients return such as
> "-4.985241"
>
> xttobit vl_share  edu_vill, ul(1) ll(0) i(countycode)
>
>                          Coef.                 =Std. Err.                z
> P>z         [95% Conf.Interval]
>
> edu_vill                -4.985241             2.46903                 -2.02
> 0.043     -9.824451    -.1460307
>
> How does one interpret the coefficient that is outside of the bounds placed
> on the model?
>
> Thank you for your guidance!
> -------------------
> Christine E. Boyle, PhD
> Scholar / Entrepreneur
> Blue Horizon Insight
> christine@bluehorizoninsight.com
> www.bluehorizoninsight.com
>
>
> *
> *   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/