Bookmark and Share

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: st: RE: storing beta coefficients for interactions in estout


From   Laura Gibson <lgibson@asc.upenn.edu>
To   "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   RE: st: RE: storing beta coefficients for interactions in estout
Date   Fri, 11 Oct 2013 14:02:16 +0000

Thanks to Joerg and Joe for their help on this problem. I now have a better understanding of how to use betas (or not to use betas) as measures of the comparability of effects in different populations. 

And just as a follow up, unlike Joe, I have been able to store "typical" information (b's and p's) from interaction terms (when they are indicated by ##) with -estout- just fine.

Cheers,
-Laura

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Joerg Luedicke
Sent: Thursday, October 10, 2013 2:54 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: RE: storing beta coefficients for interactions in estout

Remember that these -beta-s  reflect coefficients that are standardized with respect to the outcome (y) _and_ predictor variable
(x) and are calculated as beta = b*SD(x) / SD(y). For example:

sysuse auto, clear
reg price i.foreign, b
sum price foreign
di _b[1.foreign] * .4601885 / 2949.496

Now, the problem seems to be fairly obvious: since -foreign- is a binary variable, what would a standard deviation increase even mean?
In case of categorical predictors it would make more sense to standardize coefficients with respect to y only. In case of interaction effects, a similar problem arises with continuous predictors as well since you standardize with respect to the interaction term itself, which is not very meaningful either. (See the discussion here, for example:
http://www.stata.com/statalist/archive/2013-01/msg00605.html). In short, I would suggest avoiding this type of standardization altogether in case of categorical variables and/or interaction effects. That might solve the problems with -estout-, too...  :)

Joerg




On Thu, Oct 10, 2013 at 1:11 PM, Joe Canner <jcanner1@jhmi.edu> wrote:
> Laura,
>
> I'm surprised you got that far.  When I run -eststo- on a regression 
> with interactions it tells me "interactions not allowed".  This may be 
> why you are not getting the interaction betas.  (-estadd- and -estout- 
> continue to run regardless of the error.)
>
> Although the -estout- documentation doesn't seem to explicitly cover this exception, the example they use shows how to get interaction terms by multiplying the factors, something which wouldn't be necessary if it supported interactions already.
>
> Regards,
> Joe Canner
> Johns Hopkins University School of Medicine
>
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Laura 
> Gibson
> Sent: Thursday, October 10, 2013 12:28 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: st: storing beta coefficients for interactions in estout
>
> Hello,
>
> I'm enjoying using the estout package to automatically print my estimates to files - and have quickly begun to rely on the speed with which output can be transferred to Excel. I discovered today that although betas for interactions specified by ## will print in the results window, I cannot seem to find where that estimate is saved in order to add it to my stored estimate for later printing.
>
> So this works for getting the betas for the interactions:
>
> reg dv i.iv1##i.iv2, beta
>
> But if I store the estimates and estadd beta:
>
> eststo: reg dv i.iv1##i.iv2
> estadd beta
> estout, cells(beta)
>
> Only the betas for the main effects (not the interactions) are stored, and thus printed.
>
> Any ideas how to access the stored beta coefficients for the interaction?
>
> Thanks so much,
> -Laura
>
> ----------------------------
> Laura Gibson, PhD
> University of Pennsylvania
>
>
> *
> *   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/

*
*   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/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index