How to conveniently add a lot of interaction effects to a regression

[Michael Wegmuller]

I need to perform a regression and add interaction effect between at least a dozen variables.
I have to regress earnings on ethnicity, family status, gender, level of education and month dummies. All of these variables are categorical. Ethnicity is not a single variable but multiple 0/1 variables which I combined in a global for a previous regression, same for family status. I need to interact all of these variables together in a single regression.
This seems incredibly tedious to do one-by-one and I run the risk of forgetting one. Is there a convenient way to do it?

[Rich Goldstein]

I don't understand exactly what you want, but the following may help:

help fvvarlist##parens

[Maarten Buis]

That is what factor variable notation is for. To use that effectively you first need to fix your 0/1 variables: they need to be combined into a single categorical variable, e.g. a variable called marst with values 1 for never married, 2 for married, 3 for divorced, and 4 for widowed. After that you can add an interaction with a list of other variables: reg y i.marst##(i.race i.sex i.educ c.age) i.month For more see help fvvarlist