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.

Re: st: What is the effect of centering on marginal effects?

 From Lisa Marie Yarnell To "statalist@hsphsun2.harvard.edu" , "alessandro.freire@gmail.com" Subject Re: st: What is the effect of centering on marginal effects? Date Wed, 1 Aug 2012 09:56:48 -0700 (PDT)

```Thanks, Alessandro.  Centering reduces collinearity between the centered variable and an interaction term created by cross-multiplying it with another centered variable, though, right?

Yes, we had seen the "margins" command used before in the context of obtaining marginal effects for the selection equation in the heckman model.

I am still a little unclear why centering the variable would not change the beta estimated in the model, but would change the marginal effect (what we saw).

Any other guidance is appreciated.

Thanks,
Lisa

----- Original Message -----
From: Alessandro Freire <alessandro.freire@gmail.com>
To: statalist@hsphsun2.harvard.edu
Cc:
Sent: Wednesday, August 1, 2012 6:54 AM
Subject: Re: st: What is the effect of centering on marginal effects?

Dear Lisa,

Centered and uncentered models are algebraically equivalent (see
Brambor et al, "Understanding Interaction Models: Improving Empirical
Analyses" 2006). The only difference is that, in an uncentered model,
the coefficient of b1 corresponds to the marginal effect of a one unit
change in X when the conditioning variable Z is zero, while the
corresponding coefficient on the centered model gives you the marginal
effect of a change in X when Z is at its mean.

This means that centering variables will not reduce multicollinearity
on your model. I am not familiar with the mfx command, but I would
suggest you to either perform the estimation of marginal effects by
hand (see https://files.nyu.edu/mrg217/public/interaction.html) or use
the margins command, which supports factor variables, if you are using
Stata 11.

I hope this helps.

Best wishes,
Alessandro Freire

On Tue, Jul 31, 2012 at 11:53 PM, Lisa Marie Yarnell
<lisayarnell@yahoo.com> wrote:
>
> Hi Stata users,
>
> What is the effect of centering one's predictors on marginal effects? We centered our variables prior to cross-multiplication in the creation of interaction terms to avoid multicollinearity. We found that centering did not change the effect of the beta in our heckman model, but it eliminated the marginal effects produced by a mfx dyex command.  Why would that be?
>
> I don't think it's due to Stata estimating the marginal effects "at" another value because Stata estimates marginal effects at the mean by default, and the mean of the uncentered variable at hand (M = .29) is the equivalent of the mean of the centered variable (M = .00) because in centering the variable, we subtracted .29 from all scores.
>
> Thanks,
> Lisa
>
> *
> *  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/

*
*  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/

*
*   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/
```