[R-sig-Geo] sphet using two Ws in the first equation

[Roger Bivand]

On Sun, 1 Dec 2013, Antonio Bubbico wrote:

> Hello,
>
> I am wondering if it is possible to run the following model through
> lagsarlm in spdep package:
>
> y = X \beta + \rho_{1}  W_{1} y + \rho_{2} W_{2} y + u

No, there are no model fitting functions or impacts methods for this case, 
neither GMM nor ML. There are constraints on the coefficients that are not 
well established in the literature.

Roger

>
> Thanks,
>
> Antonio
>
>
>
> 2013/11/29 Roger Bivand <Roger.Bivand at nhh.no>
>
>> On Wed, 27 Nov 2013, Antonio Bubbico wrote:
>>
>>  Hello,
>>>
>>> I would like to run a "generalized spatial two stage least squares" using
>>> the sphet package,
>>> where the model is given by:
>>>
>>> y = X \beta + \rho_{1}  W_{1} y + \rho_{2} W_{2} y + u
>>> u = \lambda W u + \varepsilon
>>>
>>> Actually I don't know how to run that model in R, and obviously the
>>> problem
>>> is the second W in the first equation.
>>>
>>> Do you know if there is possibility to implement the code to run that kind
>>> of model?
>>>
>>
>> This model is not implemented in sphet or elsewhere in R packages as far
>> as I know in any form. A reference to something similar is:
>>
>> http://ideas.repec.org/p/ces/ceswps/_2356.html
>>
>> but you may know this already. The domain of a process of this kind is
>> something which has not been exhaustively studied.
>>
>> Hope this clarifies,
>>
>> Roger
>>
>>
>>> Regards,
>>>
>>> Antonio Bubbico.
>>>
>>>
>>>
>> --
>> Roger Bivand
>> Department of Economics, Norwegian School of Economics,
>> Helleveien 30, N-5045 Bergen, Norway.
>> voice: +47 55 95 93 55; fax +47 55 95 95 43
>> e-mail: Roger.Bivand at nhh.no
>>
>>
>
>
>

-- 
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no