RE: st: Testing for interaction the right way

[Amal Khanolkar <Amal.Khanolkar@ki.se>]

Hi Maarten,

Thanks for the suggestions.

When I tried to use the '##' to indicate an interaction I got the following error message:

. regress syst i.motherland2##bwtgestage_sd age byear i.conscript_office bmi height_cons i.education if multibirth==1
bwtgestage_sd:  factor variables may not contain noninteger values
r(452);

end of do-file

Have I left out something, or should it be done differently? (bwtgestage_sd is a continuous variable).

Thanks,

/Amal.

________________________________________
From: owner-statalist@hsphsun2.harvard.edu [owner-statalist@hsphsun2.harvard.edu] on behalf of Maarten Buis [maartenlbuis@gmail.com]
Sent: 30 July 2012 15:53
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: Testing for interaction the right way

On Mon, Jul 30, 2012 at 2:52 PM, Amal Khanolkar  wrote:
>  eststo: xi: regress syst i.motherland2*bwtgestage_sd age byear i.conscript_office bmi height_cons i.education if multibirth==1
> 1. I would like to know if I have included the interaction term (i.motherland2*bwtgestage_sd) in the right way?

You are much better off using Stata's factor variable notation:
i.e.leave the -xi- prefix away and use -i.motherland2##bwtgestage-

> 2. I would have expected more of the interaction tests to be significant (given the results by ethnic group in the first regression model above).

No, a model without interactions cannot tell you much about the
presence or absence of interaction effects.

> 3. I also ran the bwt-syst. BP regression 'stratified' by ethnic group. But then I assume it's not sensible to text for interaction after each stratified analysis separately as it does not make sense, since the bwt-syst. BP association is now a comparison only between one ethnic gropu and the reference group?

It is not a matter of not being sensible, it is a matter of being
impossible: An interaction effect means that you think that effects
are different between ethnic groups, i.e. it is a comparison. You
cannot compare if you only have one group, which is what you get when
you estimate separate models.

Hope this helps,
Maarten

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

*
*   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/