[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: How can I make a graphic of regression results using only significant effects?
--- Alvaro Zevallos <firstname.lastname@example.org> wrote:
> I didn't want to remove the interaction effect for the model, only
> for the graph. <snip> I think it`s interesting to focus on the
> statistically not significant interaction.. In order to do that I
> would need a graph with two parallel lines, but - postgr3 - doesn't
> give me an option.
If you want two paralel lines, you want to force the two slopes to be
equal, which means you want to constrain the difference between the two
slopes to be zero. The interaction term is the difference between the
two slopes. Constraining a parameter to be zero is the same as leaving
the variable out. So what you want is the graph of the model without
the interaction. (And maybe compare it with the model with the
> Anyway, I solved this problem changing all the values of the
> interaction variable for the mean.
Some people might object, since the interaction involves dummy
variables. Say the dummy is gender, and aprox. 50% of your sample is
female, than you would get predicted values for a person who is half
female... Anyhow it is not realy a solution to your problem: what you
want to achieve is to suppress the difference between two curves,
instead of forcing the difference not to exist, you hide the difference
by not showing the two curves but instead showing one curve for some
person that we know cannot exist.
Sorry for again being so critical (but better me than someone really
important like an advisor or referee).
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
1081 HV Amsterdam
Buitenveldertselaan 3 (Metropolitan), room Z434
+31 20 5986715
To help you stay safe and secure online, we've developed the all new Yahoo! Security Centre. http://uk.security.yahoo.com
* For searches and help try: