Plotting interaction with manually created interaction term

[Timo Wiesner]

Hello,

is it possible to plot an interaction using margins / marginsplot if the interaction term was created manually?

My model looks like this and I am well aware of the possibility to get the interaction terms with the ## notation i.e.:

mixed y c.x1##c.x3 c.x1##c.x4 c.x2##c.x3 c.x2##c.x4  || cntry: x1 x2, cov(uns) || cntryr: x1 x2, cov(uns) var base

Using this code, I get the desired results using margins / marginsplot. However, since I want to try a proposed technique, I am forced to create the interaction terms by hand.
I get basically identical results to the model before, but when doing it this way, the margins command after running the model doesn't seem to "get" the interaction..

foreach var of varlist x1 x2{
egen `var'_dm = mean(`var'), by (cntry)
replace `var'_dm = `var'-`var'_dm
gen `var'_gw = `var'_dm*x3
sum `var'_gw
replace `var'_gw = `var'_gw-r(mean)
}

mixed y x3 x1_gw x2_gw c.x1##c.x4 c.x2##c.x4  || cntry: x1 x2, cov(uns) || cntryr: x1 x2, cov(uns) var base

Is there a possibility to plot the interaction effect?

[Andrew Musau]

Use the -expression()- option of margins. But you will have to work out the expression involving the manual interaction term beforehand.