[R] Newey-West and SUR regression models

[Markus Jäntti]

Achim Zeileis wrote:
> On Fri, 26 Oct 2007, Richard Saba wrote:
> 
>> Is anyone aware of a procedure to apply Newey-West corrections for
>> autocorrelation to a SUR regression model? The SANDWICH package seems to be
>> applicable only to LM or GLM models.
> 
> No, "sandwich" is applicable to any estimating function-based model,
> provided estfun() and bread() methods are available. See
>   vignette("sandwich-OOP", package = "sandwich")
> for more details.
> 
> You can estimate SUR models with systemfit() but I don't think that the
> package currently provides such methods. Given that these are linear
> models, it shouldn't be too difficult to write such methods. If you do,
> you could maybe provide them to the "systemfit" authors.

You could use gls in package nlme to with the appropriate "weights" and 
"correlation" specifications to not only make corrections for autocorrelation 
but to estimate the model under correct specifications.

Regards,

Markus
-- 
Markus Jantti
Abo Akademi University
markus.jantti at iki.fi
http://www.iki.fi/~mjantti