On Fri, 3 Mar 2006, Al Leong wrote:
while trying out the 2SLS estimation for a
simultaneous equation model: IS-LM Model (Asteriou,
2006), I ecountered the following problem (see test
case attached)
You can see from the attached output that although
three instrument variables (M, I and M_1) are
specified, only one, I , was used in the estimation.
Your tsls command was
tsls R 0 M Y M_1 ; 0 M I M_1
Gretl reports as "Instruments" only those instruments that are not
also included as regressors, which in this case is only I. You can
verify that the estimation was performed as you intended by doing
the two stages manually:
ols Y 0 M I M_1
genr Yhat = $yhat
ols R 0 M Yhat M_1
The coefficients in the last ols are indeed the same as those using
the tsls command.
Allin Cottrell