Hello everybody, I need your help please.

I have a dummy variable which is potentially endogenous. I use it in 2 different equations. The forms are:

xtreg lny DUMMY controls, fe vce(cluster ID)
xtreg lny x1##x1##DUMMY controls, fe vce(clusterID)


Now I want to use the instrumental variable approach, but I'm not sure which commands to use. I have two possible instruments but I'm not sure how to make sure they are good instruments. If I use xtivreg the instruments are insignificant in the first stage of the regression, does that mean they are not good? My effect of the dummy in the second equation becomes insignificant too.

I read it's better to use treatreg if you have binary variables like I have, but in my normal model I use the fixed effects model, so is this comparable?

And how do I test for the dummies in the interaction terms with x1 and x1 squared?

Sorry for my basic questions, I have never worked with this before and can't find easy to understand material to my specific questions.

Thank you for your help!!