Dorothy Bridges<[email protected]>:
The main problem I see there is one of inferring causality--does an
exogenous increase in GDP cause an increase in "health" or does an
increase in "health" cause increased production or does some other
factor increase both? You might want -xtivreg- but then you need
appropriate excluded instruments...
On Tue, Sep 8, 2009 at 11:10 AM, Dorothy Bridges<[email protected]> wrote:
> Thanks so much, everyone, that answers my question. (And thanks Roger
> for pointing me to the helpful package). I have one last question: I
> am trying to understand the "elasticity" of a health indicator with
> respect to GDP (that is, how does a marginal increase in GDP affect
> the health indicator?). If I regress ln(healthindicator) on ln(gdp),
> is the coefficient on ln(gdp) an appropriate way of measuring this
> elasticity? I realize this is more of an econometrics question than a
> Stata question, but everyone has been so helpful I thought I'd might
> as well ask.
>
> Thanks again,
> Dorothy
>
> On Tue, Sep 8, 2009 at 10:52 AM, Austin Nichols<[email protected]> wrote:
>> Dorothy Bridges<[email protected]>:
>> The linear model and the log-log model make very different assumptions
>> about the data-generating process, so you should not expect to get the
>> same estimates from them. -mfx- after a linear model in particular
>> will give approximate estimates that are very questionable--note that
>> if the true model is
>> y=Xb+e
>> then the elasticity of y with respect to x (a variable, or column of
>> X) will vary as x varies.
>>
>> On Tue, Sep 8, 2009 at 10:32 AM, Dorothy Bridges<[email protected]> wrote:
>> I still
>>> don't see, then, how mfx, eyex gives me a value different than the
>>> coefficient on lnx in regress lny lnx. Thanks again for all your
>>> thoughts.
*
* 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/
