Re: st: does margins command obviate concerns that inteff solves?

[Maarten Buis <maartenlbuis@gmail.com>]

The main point of -inteff- is that it deals with marginal effects for
interaction terms in logit/probit models. If you want adjusted
predictions than you won't need it.

I thoroughly disagree with Pinar Karaca-Mandic, Edward C. Norton and
Bryan Dowd  on one point: They advise against the use of the odds
ratio interpretation. They have two arguments for that. First, odds
ratios aren't risk ratios. I have heard that complaint before, but to
me it just sounds like complaining that a meter is a bad unit because
it is a bad approximation of a kilogram. When you report odds ratios
than you should just interpret them as a ratios of odds, because that
is what they are. Complaining that odds ratios aren't something else
just doesn't make sense to me.

Second, the odds ratio interpretation is often misinterpreted, and
that that is a reason for discouraging the odds ratio interpretation.
That would be correct if odds ratios where inherently hard, but that
is a myth. People just aren't used to them. So, you have to help your
audience a bit with the interpretation, and that is really easy and it
fits naturally within the standard form of an academic paper. I have
given an example in:
M.L. Buis (2010) "Stata tip 87: Interpretation of interactions in
non-linear models", The Stata Journal, 10(2), pp. 305-308.

-- Maarten

On Fri, Dec 9, 2011 at 4:47 PM, Doug Hess wrote:
> Hi.
>
> I have been searching around but couldn't find confirmation that the
> -margins- command handles well the concerns about interactions in
> logistic regression that led to the user-written command -inteff-.  I
> see an article forthcoming in the Health Services Research that might
> cover this, but I cannot access it yet (by Pinar Karaca-Mandic, Edward
> C. Norton and Bryan Dowd; to find this article, google:  DOI:
> 10.1111/j.1475-6773.2011.01314.x  ). I have seen some mention that
> -margins- might obviate the need for using inteff, but it was not
> certain. Finally, there are comments that using relative risk ratios
> gets around this, etc., but I am concerned with adjusted predictions
> following interactions and using the -margins, contrast - options or
> contrast operators after estimating a model with interactions, and I'd
> like to make sure that I'm not falling into the trap that Norton, et
> al. have mentioned in many articles (
>
> Feel free to cc me on responses at douglasrhess@gmail.com, as I get
> the list as a digest and will not see responses quickly (i.e., to be
> able to respond or follow up, etc.).
>
> Thank you.
>
> Doug
>
> *
> *   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/



-- 
--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany


http://www.maartenbuis.nl
--------------------------

*
*   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/