What is wrong with my code on instrumented DID regression "xtivreg y (i.shock#i.treatment = i.shock#iv) i.year c.year#id x, fe"?

[Linghui Han]

I run the code "xtivreg y (i.shock#i.treatment = i.shock#iv) i.year c.year#id x, fe vce(cluster id)" but got the error "depvars may not be interactions". In a DID setting, is it ok to have an instrument variable with continuous values instead of binary values?

Thanks in advance!

[Linghui Han]

Or what is the code for instrumented DID regression with panel data?

[Linghui Han]

Is my question clear? I am wondering how to make it clearer for others who might be able to help.

[Linghui Han]

Problem solved: replace those interactions with a new variable (e.g., gen shock_treatment=i.shock#i.treatment).

But what is the code for the first stage regression? I have not figured that out yet.