Synergistic effects of three independent variables

[Guest]

Hey there,

I'm doing research on whether three variables together have more effect on dependent variable than either by themselves. In other words:
Effect on Y if we only have X1
Effect on Y if we only have X2
Effect on Y if we only have X3
Effect on y if we have both X1, X2, X3

Initially, i was comparing:
y = x1 x2 x3
vs.
y = Z with Z=X1+X2+X3
--> where i just interpreted beta X1 vs. beta Z, to see whether the effect was stronger or not.

I just found out that this is wrong, though i don't understand how to remedy this. my deadline is in a couple of days and my promoters never told me that it was wrong. So, what exactly do i have to do then? I created a new variable by multiplying each separate variable, but that doesnt seem to fit either.

note: dependent and all independent variables are ordinal

[Carlo Lazzaro]

Guest:
without other details, it would seem a matter of interaction.
Assuming you have a continuous dependent and independent variables (if your independent variables were categorical, just replace the prefix -c.- with the prefix -i.-:

reg Y c.X1##c.X2##c.X3

See more on categorical variables and interaction creation in -help fvvarlist-.
Consider -margins- and -marginsplot- as relevant supports for explaining your regression outcome to your audience (even graphically).

[Bruce Weaver]

The following UCLA web-pages will help you understand Carlo's advice in #2.

• http://stats.idre.ucla.edu/stata/faq...tion-stata-12/
• http://stats.idre.ucla.edu/stata/faq...nova-stata-12/

HTH.

[Guest]

Many thanks, that was really helpful. However, my variables display high collinearity, which makes it impossible to calculate. Should i just mention this in the paper ? (independent variables have > 0.9 correlations). I tried centering mean etc.. but nothing's successful. So, is it reasonable to just mention something like that, or try something different?