In your logit model, the only information provided by the coefficient
on your interaction term is whether it belongs into the specification,
that's it!
For any further effect, you need to bootstrap your estimation or use
ado file based on the Delta Method mentioned in the web pages
mentioned below by Marteen.
France
On 4/21/05, Maarten Buis <[email protected]> wrote:
> Dear Ying,
>
> When I use interactionterms within logistic regresion, I use the standard way of dealing with interactionterms on the output of -logit- itself before any other computation, like computing odds ratio's or marginal effects. This is what I meant with "raw" output. Since I often use many and/or complicated interactionterms I often do the computations by 'hand', well write code like those in the threads I refered to earlier, and not use -mfx- or the like.
>
> Maarten
>
> -----Original Message-----
> From: [email protected] on behalf of [email protected]
> Sent: Thu 4/21/2005 8:03 PM
> To: [email protected]
> Cc:
> Subject: Re: st: RE: interpret the marginal effect of interaction term in logit model
> Dear Maarten,
>
> I am looking into the threads you provided. Just want to clear things a
> little bit. When you say "yes" when B1 and B3 are "raw" output, what do
> you mean "raw"? Is "mfx compute" relevant or not?
>
> Ying
>
> > Dear Ying,
> >
> > The answer is yes if B1 and B3 are 'raw' output from your logit model,
> > otherwise the answer is no.
> >
> > Have a look at:
> > http://www.stata.com/statalist/archive/2005-03/msg00643.html . Also, a
> > similar question (about getting marginal effects for a quadratic term
> > instead of an interaction term) was asked last year and got some useful
> > answers (if I say so myself). The thread started with:
> > http://www.stata.com/statalist/archive/2004-11/msg00254.html
> >
> > Hope this helps,
> > Maarten
> >
> > -----Original Message-----
> > From: [email protected]
> > [mailto:[email protected]]On Behalf Of [email protected]
> > Sent: donderdag 21 april 2005 7:36
> > To: [email protected]
> > Subject: st: interpret the marginal effect of interaction term in logit
> > model
> >
> > Dear listusers,
> >
> > I have a quick question about the marginal effect of logit model.
> >
> > Suppose I put black, female, and the interaction term of black and female
> > (blackfem) in the logit model, and use "mfx compute" to obtain the
> > marginal effects. My dependent variable is a dichotomous variable:
> > person's working status. My model is:
> >
> > working status (Y) = B0 + black*B1 + female*B2 + blackfem*B3 + ... +
> > error term.
> >
> > Regardless of the significance value, can I simply interpret B1-B3 as
> > what we do for OLS? If not, how did you interpret them in English?
> >
> >
> > Thanks in advance,
> > Ying
> >
> >
> >
> > *
> > * For searches and help try:
> > * http://www.stata.com/support/faqs/res/findit.html
> > * http://www.stata.com/support/statalist/faq
> > * http://www.ats.ucla.edu/stat/stata/
> >
> > *
> > * For searches and help try:
> > * http://www.stata.com/support/faqs/res/findit.html
> > * http://www.stata.com/support/statalist/faq
> > * http://www.ats.ucla.edu/stat/stata/
> >
>
> Ying Wang
> Ph.D. Candidate
> Department of Public Policy
> University of Maryland, Baltimore County
> 1000 Hilltop Circle,
> Baltimore, MD 21228
> Tel(O): 410-455-6531
> Tel(H): 410-747-6980
> Fax: 410-455-1172
> Email: [email protected]
>
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
>
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
