I have a question concerning 2SL2 estimation with several instrumented variables.
my idea is to do the following
ivreg y k (x= d f) (z= r g), instrumenting x with d and f and z with r and g
however as far as I understand stata, the program suggests to do the following:
ivreg y k (x z= d f r g), meaning that I have to instrument x and z with the same variables d f r g.
An other way could be to estimate the predicted x and z and then to use the predicted values in the second step.
Then I had to correct for the standar errors.
My problem is that in fact I want to instrument not only 2 varaibles( x and z), but 24 variables.