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

Re: st: Question on interpreting interactions in linear regression models

From   Maarten buis <>
Subject   Re: st: Question on interpreting interactions in linear regression models
Date   Tue, 15 Dec 2009 08:03:20 -0800 (PST)

--- On Tue, 15/12/09, Padmakumar Sivadasan wrote: 
> I was reading a paper in which the authors are interested
> in studying whether the effect of a continuous variable 
> (X) on a continuous outcome variable (Y) varies between
> two periods (P=0 and P=1). They test this using interactions
> with the following linear regression model  Y=b1+b2X+b3P+b4XP
> They find that b2 is significantly different from zero,
> while b4 and (b2+b4) are not significantly different from
> zero. The authors conclude that X has an influence on Y in
> the first period (i.e. when P=0) because b2 is significant.
> Further, they state that the influence of X in the second
> period (i.e. when P=1) is given by (b2+b4) and this is not
> significantly different from zero, implying that X has no
> influence on Y in the second period. Thus, the effect of X
> on Y varies between the two periods.

If the authors want to say something about the change, then he 
should look at the interaction effect alone. The fact that a 
effect is significant in one period and insignificant in another
is in itself no proof that a change has occured. See for example
this article from The American Statistician:

Hope this helps,

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


*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index