How to visualize the marginal effect of interaction term?

[Karel Novak]

Dear all, I would like to ask you how can you visualize "the marginal effect" of two variables which were in interaction. I am using screen shot from article of Brambor2005:

Which is based on this reg. model:

I am hundred percent positive It can be done in Stata, but I was not able to find and recommendation or guide how to do so. I am glad for any help.

[Maarten Buis]

sysuse nlsw88, clear
gen byte black = race == 2 if race <=2
label define black 0 "white" 1 "black"
label value black black
reg wage i.black##c.grade ttl_exp i.south i.union
margins, dydx(black) at(south=0 union=0 ttl_exp=10) over(grade)
marginsplot, yline(0)

[Karel Novak]

Thank you for your answer, but to be honest, I am unable to read the code properly. Can you elaborate a bit on which of those words can, or better, should be rewritten, so I can adapt it to my dataset? Thanks.

[Maarten Buis]

The first line opens an example dataset that comes with Stata

line 2-4 create a new variable for race distinguishing only between blacks and whites

line 5 estimates the regression with interaction effect between black and grade

line 6 computes the effect of black for different values of grade

line 7 plots this in a graph

You may want to look at help fvvarlist , help margins , and help marginsplot

[Karel Novak]

Thanks a lot, I will surely look at the urls (edit: help commands), but just to be sure, one quick question. I have only scale variables, is the code same for this situation?

[Maarten Buis]

For continuous variables you use the c. prefix (like c.grade) and for categorical variables you use the i. prefix (like i.black). See help fvvarlist.

[Maarten Buis]

In general, if you want us to provide code that fits your problem, you need to provide us with example data that fits your problem. Otherwise, we are forced to use generic example data. See the FAQ on how to provide such example data.