Interpreting nbreg output with interaction terms

[Arnisa Kastrati]

Hello,

I am having trouble interpreting this output. Would someone be able to provide advise?

The interaction terms are: issue type (migration, environment,energy, other) x organization type (business, NGO, other)

[Bruce Weaver]

Hi Arnisa. It appears that you computed your own indicator variables and product terms. If you use factor variable notation instead (with # or ## for the interaction), you'll be able to use -margins- and -marginsplot- to visualize the nature of the interaction.

help factor variables
help nbreg postestimation

HTH.

[Arnisa Kastrati]

Thank you for your help. I appreciate it. However, I am still not quite sure on how to fix it.
The command I used was : xi: nbreg y1 i.issue1*i.orgtype_cat3 lobby budget

when using # for the interaction, I get back :

interactions not allowed
r(101);

[Clyde Schechter]

You don't use factor variable notation with -xi-, you use it instead of -xi-.

nbreg y1 i.issue1##i.orgtype_cat3 lobby budget

Note also that for a well specified model you have to have both the interaction term and each of its constituents included. This happens automatically when you use the ## notation. If you used # instead of ## you would have to also explicitly include i.issue1 and i.orgtype_cat3 in the variable list.

Factor variable notation is supported by all modern Stata estimation commands. Some very old commands will not allow it, and there are a few odd situations where -xi- is still needed even in modern estimation commands. But those are rare. You should almost forget you ever heard of -xi- and just use factor variable notation going forward.

[Arnisa Kastrati]

Thank you, it was really helpful.