Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Re: st: Error w/ "inteff" command
From
Erasmo Giambona <[email protected]>
To
[email protected]
Subject
Re: Re: st: Error w/ "inteff" command
Date
Wed, 9 Jan 2013 21:23:41 +0100
Thanks a lot Rich. After more thinking and the additional readings
that you suggested, my feeling is that one should be able to use the
slope coefficient for the interaction term in case of LPMs.
Erasmo
On Wed, Jan 9, 2013 at 2:00 AM, Richard Williams
<[email protected]> wrote:
> At 12:36 PM 1/8/2013, Erasmo Giambona wrote:
>>
>> Dear Richard,
>>
>> I have read your Stata's article carefully and I think I understand
>> better why marginal effects on interaction terms (e.g., interactio of
>> two dummy variables) do not exist in the case of non-linear models.
>> But what about Linear Probability Models? Can I simply use the slope
>> coefficient on the interaction term in this case to measure the
>> marginal effect? I would appreciate you answer on this issue.
>>
>> Best regards,
>>
>> Erasmo
>
>
>
> First off, I shouldn't have said there was no marginal effect for an
> interaction term, because there is. But, I'll still stick with saying I
> don't think it is very useful and that there are better ways to examine
> interactions. Look at these recent posts from StataCorp people in case you
> have missed them:
>
> http://www.stata.com/statalist/archive/2013-01/msg00263.html
>
> http://www.stata.com/statalist/archive/2013-01/msg00293.html
>
> As for your question -- off the top of my head I don't know. Maybe somebody
> else does, or maybe you could easily figure it out by adapting examples that
> Vince Wiggins gives in the 2nd post above. Rich
>
>
>> > On Wed, Jan 2, 2013 at 6:22 PM, Richard Williams
>> > <[email protected]> wrote:
>> >> I don't really understand how -inteff- works, nor do I have any great
>> >> desire
>> >> to find out. I am happy with the -margins- command, and the way you set
>> >> it
>> >> up is not correct for -margins-. When you compute the interaction term
>> >> yourself, Stata has no way of knowing that the values of the
>> >> interaction
>> >> term are not independent of the values of the variables used to compute
>> >> it.
>> >> It should be
>> >>
>> >> webuse lbw2
>> >> probit low age lwt c.age#c.lwt
>> >> margins, dydx(_all)
>> >>
>> >> For an explanation, see
>> >>
>> >> http://www.nd.edu/~rwilliam/stats/Margins01.pdf
>> >>
>> >> or else
>> >>
>> >> http://www.statajournal.com/article.html?article=st0260
>> >>
>> >>
>> >> At 10:48 AM 1/2/2013, Erasmo Giambona wrote:
>> >>>
>> >>> Dear Kit,
>> >>>
>> >>> I was finally able to get the "inteff" command to work again. Inteff
>> >>> and margins give me estimates on the interaction term that are
>> >>> similar, but not the same. Is this simply do to different
>> >>> approximation? Thanks. Please, see example below (using: webuse lbw2):
>> >>>
>> >>>
>> >>> . g age_lwt=age*lwt
>> >>>
>> >>> . probit low age lwt age_lwt
>> >>>
>> >>> Iteration 0: log likelihood = -117.336
>> >>> Iteration 1: log likelihood = -113.61015
>> >>> Iteration 2: log likelihood = -113.58509
>> >>> Iteration 3: log likelihood = -113.58509
>> >>>
>> >>> Probit regression Number of obs =
>> >>> 189
>> >>> LR chi2(3) =
>> >>> 7.50
>> >>> Prob > chi2 =
>> >>> 0.0575
>> >>> Log likelihood = -113.58509 Pseudo R2 =
>> >>> 0.0320
>> >>>
>> >>>
>> >>>
>> >>> ------------------------------------------------------------------------------
>> >>> low | Coef. Std. Err. z P>|z| [95% Conf.
>> >>> Interval]
>> >>>
>> >>>
>> >>> -------------+----------------------------------------------------------------
>> >>> age | -.0316919 .0896229 -0.35 0.724 -.2073495
>> >>> .1439658
>> >>> lwt | -.0087146 .0162868 -0.54 0.593 -.040636
>> >>> .0232069
>> >>> age_lwt | .0000561 .0006749 0.08 0.934 -.0012666
>> >>> .0013788
>> >>> _cons | 1.186736 2.124989 0.56 0.577 -2.978165
>> >>> 5.351637
>> >>>
>> >>>
>> >>> ------------------------------------------------------------------------------
>> >>>
>> >>> . inteff low age lwt age_lwt
>> >>> Probit with two continuous variables interacted
>> >>> (0 observations deleted)
>> >>>
>> >>> Variable | Obs Mean Std. Dev. Min Max
>> >>> -------------+--------------------------------------------------------
>> >>> _probit_ie | 189 .0000473 7.14e-06 .0000265 .0000548
>> >>> _probit_se | 189 .0002247 .0000631 .0000304 .0002841
>> >>> _probit_z | 189 .2615582 .2009778 .1001468 1.322986
>> >>>
>> >>> . margins, dydx(_all)
>> >>>
>> >>> Average marginal effects Number of obs =
>> >>> 189
>> >>> Model VCE : OIM
>> >>>
>> >>> Expression : Pr(low), predict()
>> >>> dy/dx w.r.t. : age lwt age_lwt
>> >>>
>> >>>
>> >>>
>> >>> ------------------------------------------------------------------------------
>> >>> | Delta-method
>> >>> | dy/dx Std. Err. z P>|z| [95% Conf.
>> >>> Interval]
>> >>>
>> >>>
>> >>> -------------+----------------------------------------------------------------
>> >>> age | -.0108404 .0306064 -0.35 0.723 -.0708278
>> >>> .0491469
>> >>> lwt | -.0029809 .0055547 -0.54 0.592 -.0138679
>> >>> .0079061
>> >>> age_lwt | .0000192 .0002308 0.08 0.934 -.0004332
>> >>> .0004715
>> >>>
>> >>>
>> >>> ------------------------------------------------------------------------------
>> >>>
>> >>>
>> >>>
>> >>>
>> >>>
>> >>>
>> >>>
>> >>>
>> >>> On Sat, Dec 29, 2012 at 11:16 PM, Christopher Baum <[email protected]>
>> >>> wrote:
>> >>> > <>
>> >>> > Erasmo said
>> >>> >
>> >>> > Does - margins, dydx(_all) - also handle the interaction of two
>> >>> > dummy
>> >>> > variables?
>> >>> >
>> >>> > Yes. ht (hypertension, yes/no) and smoke (yes/no) are such, and
>> >>> > interacted in the model below. Notice that each has a positive main
>> >>> > effect
>> >>> > on low bw, but if they appear together the effect is, strangely
>> >>> > enough,
>> >>> > reduced (although the negative interaction coefficient is not
>> >>> > distinguishable from zero).
>> >>> >
>> >>> > probit low c.age##i.race i.ht##i.smoke
>> >>> > margins, dydx(_all)
>> >>> >
>> >>> >
>> >>> >
>> >>> > Kit Baum | Boston College Economics & DIW Berlin |
>> >>> > http://ideas.repec.org/e/pba1.html
>> >>> > An Introduction to Stata Programming |
>> >>> > http://www.stata-press.com/books/isp.html
>> >>> > An Introduction to Modern Econometrics Using Stata |
>> >>> > http://www.stata-press.com/books/imeus.html
>> >>> >
>> >>> >
>> >>> > *
>> >>> > * 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/
>> >>
>> >>
>> >> -------------------------------------------
>> >> Richard Williams, Notre Dame Dept of Sociology
>> >> OFFICE: (574)631-6668, (574)631-6463
>> >> HOME: (574)289-5227
>> >> EMAIL: [email protected]
>> >> WWW: http://www.nd.edu/~rwilliam
>> >>
>> >> *
>> >> * 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/
>
>
> -------------------------------------------
> Richard Williams, Notre Dame Dept of Sociology
> OFFICE: (574)631-6668, (574)631-6463
> HOME: (574)289-5227
> EMAIL: [email protected]
> WWW: http://www.nd.edu/~rwilliam
>
> *
> * 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/