I would need your help with the following.
I have two models
1) Rf = a + b1*P
2) Rf = a + b1*P + b2*R + b3*I + b4*S + b5*V + b6*C + b7*A + b8*E + b9*G
I have check done all diagnostics for autocorrelation, heteroscedasticity, etc and they both seem fine.
The problem is that variable P appears in both equations and it may be endogenous in the system with the dependent variable Rf.
Is the J-test (Davidson-MacKinnon) the right way to check this or not and if yes how do I do it?
Is it that I run the first regression get the fitted values and the do the same for regression 2? And then do I plug in the fitted values in the regression by removing the variable P from the second regression?
Another question I have is the following:
When running 2STLS I leave variable P in the regressors list with the rest of the variables of eq2, while in the instruments lists I insert the rest of variables (variables R,I,S,C,A,E,G). The model will not run. Why is that?
Furthermore, do I insert Rf also in the instruments list?
Finally, if the instruments are right or wrong, or lets lets if I do not need a 2STLS and the OLS works fine, how do I check that?