Dear Leny,
Excluding a interactionterm means that the trend is forced to be the
same for group one and two. This is not a problem, if the trends are
indeed close to equal. However, if this is not the case, regression
will still try to find the best fitting curve given your constraints.
This could lead to drastic differences between the interaction and
non-interaction model. You could make a scatterplot and graph the
predicted regressionline for group one and two seperately, for the two
models and see if the trend in group one is indeed different from
group two. (i.e. perform the 'interocular trauma test').
Hope this helps,
Maarten
--- <lm335@d...> wrote:
>
> Dear Statalisters,
>
> I have two regression models:
> 1) ln(y)= constant + b1*(time) + b2*(group)
> 2) ln(y)= constant + b1*(time) + b2*(group) + b3*(group*time)
> How does one explain (In plain English) the drastic reduction in the
>effect of time on y for the group2 people between the two models?
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
