Interaction questions

[Omar Shaher]

Dear researchers,
I have quick questions on the interaction methods, please:
I have unbalanced panel data set for a set of firms for the years extends from 2010-2015.
I have the following model:

Dep = X1 + X2 + X3 +X4

I have used firm and year fixed effect regression.
Q1)
I want to interact X3 with X4 and both of these variables are continuous variables. I don’t know if this method called “ multiplicative method for interaction” or not. I hope you correct me?
Q2) Anyway, to do the interaction, which method I should use:

Dep = X1 + X2 + X3
Dep = X1 + X2 + X3 +X4
Dep = X1 + X2 + X3 +X4 + X3*X4

OR I should go directly with:

Dep = X1 + X2 + X3 +X4 + X3*X4

Q3) How to interpret the interaction for the above two continuous variables. Shall I look to the coefficient before the interaction and sum it to the coefficient of the interaction to know what happened like the DID method or what?
Q4) Can I get the margin for the above and the graph to see what happened after the interaction? If yes, please what is the code?

[Oscar Ozfidan]

the idea is correct with
Dep = X1 + X2 + X3 +X4 + X3*X4 representation
but

Dep = X1 + X2 + X3 +X4 + X3*X4 wont work.


reg Dep X1 X2 X3 X4 c.X3#c.X4 would work in Stata

as far as interpretation, you can plot predicted Dep when X3 is low (say 25th percentile) and draw a line between two points one for predicted Dep when X4 is low and the second point when X4 is high
then do the same for when X3 is high (say 75th percentile). You will have four points and two lines and that should give you an idea on what the interaction term does.

I dont know about the margins command and cant help you with Q4

[Maarten Buis]

Here is how you would use margins for what Oscar Ozfidan proposed: http://www.maartenbuis.nl/wp/inter_q...quadr_new.html

[Omar Shaher]

Thanks, Oscar and Maarten, now the issue is clear for me. Thank you a lot.