RE: st: Bug in margins

[Richard Williams <richardwilliams.ndu@gmail.com>]

At 10:58 AM 8/30/2013, Joe Canner wrote:
> Roland,
>
> 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).  The original model runs fine because binary 
> variables don't need the "i.", but for whatever reason -margins- requires it.

With the margins command, the vars to the left of the comma must be a 
"list of factor variables or interactions that appear in the current 
estimation results." The i. notation in the estimation command tells 
Stata the variable is categorical; otherwise it assumes it is 
continuous. When you put a factor variable on the left hand side of 
the margins command, Stata is going to do things like give you 
predicted values for men and women. You don't want to put a 
continuous variable on the left hand side because continuos variables 
potentially have an infinite number of values.

The continuous /categorical distinction is also important because 
marginal effects are computed differently for continuous and categorical vars.

Of course, you might say, if a variable is coded 0/1, why doesn't 
Stata just assume it is categorical rather than continuous? After 
all, that is what the old -mfx- command does. I believe Nick Cox made 
the point once that a variable coded 0/1 could actually be 
continuous, but only the values of 0 and 1 were observed in the 
sample. Plus I suppose it makes it easier for coding if Stata 
requires you to be explicit about whether or not a variable is categorical.

> 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 Richard Williams
> Sent: Friday, August 30, 2013 12:51 PM
> To: statalist@hsphsun2.harvard.edu; statalist@hsphsun2.harvard.edu
> Subject: Re: st: Bug in margins
>
> First off, your 2 logistic commands look identical to me. I'm 
> guessing that when you ran it the first time and got the error, you 
> did not have the i.
>
> Assuming that is the case, this is no error. The i. tells Stata that 
> the variable is categorical rather than continuous. See the help for 
> margins. Or, for some highlights on the margins command, see
>
> http://www3.nd.edu/~rwilliam/stats/Margins01.pdf
>
>
>
> At 10:39 AM 8/30/2013, roland andersson wrote:
> >I am analysing a logistic regression model, binary outcome. I want to
> >analyse the adjusted proportion of the outcome for a binary (0/1)
> >covariate.
> >
> >I run the logistic model followed by margins . logistic dod30dgr
> >i.lapscopiintent alder aldersq........
> >. margins lapscopiintent
> >
> >I get an error
> >'lapscopiintent' not found in list of covariates r(322);
> >
> >If I put i. before the binary variable I am interested in
> >
> >. logistic dod30dgr  i.lapscopiintent alder aldersq................
> >. margins lapscopiintent
> >
> >Predictive margins Number of obs   = 166811
> >Model VCE    : OIM
> >
> >Expression   : Pr(dod30dgr), predict()
> >
> >Delta-method
> >Margin   Std. Err. z P>z [95% Conf. Interval]
> >lapscopiin~t
> >0     .0015466   .0000985 15.70 0.000 .0013535 .0017397
> >1     .0011487   .0002518 4.56 0.000 .0006552 .0016421
> >
> >I assume this is a bug?
> >
> >Roland Andersson

-------------------------------------------
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/