Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: question about the interaction term


From   ZHVictor <victerzj2@hotmail.com>
To   "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   RE: st: question about the interaction term
Date   Thu, 25 Apr 2013 08:33:54 +0000

----------------------------------------
> Date: Thu, 25 Apr 2013 10:12:08 +0200
> Subject: Re: st: question about the interaction term
> From: maartenlbuis@gmail.com
> To: statalist@hsphsun2.harvard.edu
>
> On Thu, Apr 25, 2013 at 9:32 AM, ZHVictor wrote:
> > So that means whenever I have the similar regression, I should use "test A+A*B=0" to double check, rather than only look at the interaction term.
>
> No, that is exactly opposite of what the main point of that article is.


OK, I am thinking the basic idea of the paper is even the small and insignificant change can lead to a large change in significance level. Applying it to my example, it means even the interaction term is insignificant, the coefficient of A may be insignificant in B=1 case even thought it is significant in B=0 case. Thus, if I only look at the insignificant interaction term, I may conclude that the coefficient of A is still significant in B=1 case as in B=0 case.  So it means even though the interaction term is insignificant, it may still be able to change the significance level of A. 


>
> > Thus, for B=0 case, I should only look at the p-value of the coefficient of A to see whether the coefficient of A is significant.
> > However, for B=1 case, I should actually test whether A+A*B is significant (use test A+A*B=0). If A+A*B is insignificant different from zero, I should say A has on effect on Y when B=1, even if the interaction term is insignificant.
> > Is what my understanding correct?
>
> No, the trick is to work out exactly what the null hypothesis is that
> you want to test and create a test that tests exactly that null
> hypothesis. An interaction term measures exactly what an interaction
> term measures and if that is what you want to know then that is
> enough.

For my case, I need to know two things. First, whether the coefficient of A is significant in B=0 case. Second, whether the coefficient of A is significant in B=1 case. Thus, for the B=0 case, it is easy to see the significance of A in the regression. However, for B=1 case, the coefficient of A becomes the coefficient of A+the coefficient of A*B. Therefore, I should test A+A*B=0. Right?


>
> > One more question is if the coefficient of A is -0.4 and the coefficient for the interaction is 0.2, so the coefficient of A in B=1 case should be -0.4+0.2=-0.2 but not -0.4+0=-0.4. Is that correct?
>
> I don't understand that question, where did the 0 come from?

Since the interaction term is insignificant, I can treat it as 0. That is where the 0 comes from.


>
> -- Maarten
>
> ---------------------------------
> Maarten L. Buis
> WZB
> Reichpietschufer 50
> 10785 Berlin
> Germany
>
> http://www.maartenbuis.nl
> ---------------------------------
> *
> * 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/ 		 	   		  

*
*   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/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index