RE: st: Bug in margins

[Joe Canner <jcanner1@jhmi.edu>]

In the PDF documentation, even StataCorp recognizes the ambiguity:

"Concerning 1, when you specify the estimation command, covariates that are logically factor variables must be Stata factor variables."

But unless one heeds the warning in -help margins- to look at the Remarks section of the PDF documentation one may not be aware that the error message is referring to the adding of "i." in the original model.

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Richard Williams
Sent: Friday, August 30, 2013 2:27 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: Bug in margins

At 11:53 AM 8/30/2013, you wrote:
>On Aug 30, 2013, at 10:58 AM, Joe Canner <jcanner1@jhmi.edu> wrote:
> > The requirement that the original model be specified including
> "i." is also buried deep in the PDF documentation for -margins- (page 
> 1063 of r.pdf).
>
>
>Actually, the very first line displayed by -help margins- is
>
>         margins [marginlist] ...
>
>     where marginlist is a list of factor variables or interactions that
>     appear in the current estimation results.
>
>A pedant might argue that this should read "a list of factor variables 
>or *their* interactions," but regardless, the information
>*is* accessible.
>
>That said, the fact that so many ask this question does perhaps suggest 
>that a more prominent note be added to the documentation and/or that 
>the error message in this case be changed to "only factor variables and 
>their interactions are allowed," which is the error message you would 
>have gotten if you had typed
>
>     margins c.lapscopiintent

I like your error messages better. Other possible sources of confusion:

. margins gpa
gpa:  factor variables may not contain noninteger values

True, but gpa wasn't a factor variable to begin with, so it would be better to give the "only factor variables" message.

Even ""a list of factor variables or *their* interactions," is potentially confusing because you can't include i.gender#c.income, i.e. the interactions can only be among the factor variables.

And, for good measure, the interactions don't have to be in the estimation command, e.g.

reg y i.x1 i.x2
margins x1#x2

is legitimate. There is no interaction term in the model but if, say,
x1 and x2 both had 2 values, you would get a 4 part classification. 
So I am not even sure what "their interactions" means. The margins notation looks like interactions but no interactions are needed in the model.

I am sure it is hard to come up with crystal clear explanations for every error. I think the biggest thing would be to get the "only factor variables" message out there more often. The "covariates" 
wording will be very unclear to many people; if you say "factor variables" people have more of a fighting chance of figuring out what the problem is.




-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
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/