Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# Re: st: question about the interaction term

 From David Hoaglin To statalist@hsphsun2.harvard.edu Subject Re: st: question about the interaction term Date Thu, 25 Apr 2013 06:25:23 -0400

Dear Vic,

You have a single model, not separate "cases" for B=0 and B=1.

You have not included any output, so I am not able to work with the
numerical estimates of the coefficients.

According to the test that you mentioned, when B=1, the slope against
A is "like 0."  I suspect that, when the coefficient of A*B is added
to the coefficient of A, that sum is close enough to 0 that it is not
significant.

It may help to remember that, as in any regression model, the
definition of each coefficient includes the set of other predictors in
the model.

You may want to remove the interaction from your model and see what
happens to the coefficient of A (which does not have the same
definition as the coefficient of A in the model that includes the
interaction).

David Hoaglin

On Thu, Apr 25, 2013 at 12:12 AM, ZHVictor <victerzj2@hotmail.com> wrote:
> Dear David,
> Thank you for your response. I understand what you are saying.
> In B=0 case, I have a significant slope for A
> In B=1 case, the slope of A becomes coefficient of A+coefficient of A*B, however it becomes insignificant.
> A*B is an interaction term. I have an insignificant coefficient of A*B. That means the coefficient of A*B is like 0
> Thus, in B=1 case, the slope of A is like coefficient of A+0. Therefore, I am expecting in B=1 case, the slope of A should be also significant as in B=0 case.
> That is where I am confused.
> Thank you!
*
*   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/