Re: [Gretl-users] instrumental variables and gauss newton regression
Hi Robert,
Robert Mathew schrieb:
1) if my model was something like the following: Mt = B1 + B2*Rt +
B3*Yt + B4*Mt-1 + B5*Mt-2 + u - how would I perform a durbin wu
hausman test for the hypothesis that my variable Rt is exogenous, if
I use Rt-1 and Rt-2 as additional instruments
Model -> instrumental variables -> two-stage least squares, and then you
specify Rt, Yt, Mt-1, Mt-2 as independent variables, and Yt, Mt-1, Mt-2,
Rt-1, Rt-2 as instruments. The Hausman test is printed out automatically
and in this case then refers only to Rt (the only instrumented rhs
variable).
- also, how would I estimate the above equation by generalized
instrumental variables, treating Rt as endogenous, and Rt-1 and Rt-2
as additional instruments?
I'm not sure what you mean by "generalized" (maybe you mean as opposed
to the special case of indirect least squares?), but for the normal
instrumental variables estimator see above
2) For a non-linear model such as: y = B1*x1 + B2*x2 + B3*x3^g + u
-how can I use a succession of Gauss Newton Regressions to obtain the
NLS estimates of all the parameters and their standard errors?
Model -> nonlinear models -> nonlinear least squares gives you an
admittedly spartan input window, but the gretl manual is your friend
here. Part II chapter 16. If anything remains unclear, feel free to ask
again.
good luck,
sven