Hello,

I can't show my data or real variable names but I have a question about whether what I want to accomplish is doable through a 2SLS regression analysis or through another way.

I ran a linear regression of VAR1= a+B1.VAR2+controls+e, the coefficient of the variable of interest on VAR2 is negative and significant.
on my second regression I rand VAR3= a+ C1.VAR2+ C2.VAR2*VAR4+C3.VAR4+ controls+s the coefficients on the VAR2 and VAR4 are negative and on the interaction term is positive.

My question is how to test the impact of decreasing VAR1 due to the occurrence of VAR2 on VAR3 through an additional term in the interaction term between VAR2*VAR4?
So ideally what it would mean is testing VAR2*VAR4*(VAR1 conditional on VAR2=1) on VAR3.

I think a three way interaction term might not be the right solution because of multicollinearity issues.
then I thought maybe I should run the first regression and use the predicted value of VAR1 and insert it in the second regression, as what I would do in a 2SLS regression analysis, but I am not sure it would actually test what I am looking for to test: does a decrease in VAR1 due to VAR2 strengthens or weakens the coefficient on the interaction term between VAR2 and VAR4 in the second regression.


Thank you for taking the time to read