commons-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Phil Steitz <phil.ste...@gmail.com>
Subject [math]Re: How to use GLSMultipleLinearRegression
Date Fri, 04 Sep 2009 10:47:44 GMT
Bharani wrote:
> I have got 3 independent variable and 1 dependent variable. Size of the
> vector (n) is 200. I am trying to use GLSMultipleLinearRegression for
> regression but i am not able to figure out how to get the covariance matrix
> ( Omega)
> 
> Thanks
> Bharani

That depends on your assumptions on the error structure of your
model.  If you believe that the errors in your model are constant
variance and uncorrelated, then you should just use ordinary least
squares (implemented by the OLSRegression class).  Otherwise, the
right choice of Omega depends on what you want to assume about how
the errors are (jointly) distributed.  See [1] for a concise
statement of the assumptions of OLS and the definition of Omega and
[2] for some common examples of assumptions and associated error
covariance matrices.  Commons math does not currently support FGLS
in the sense of [1] (i.e., we do not provide automatic estimation of
Omega for the common assumptions described in [2]).

Phil

[1] http://jackman.stanford.edu/papers/gls.pdf

[2] http://people.virginia.edu/~sns5r/classes/grad/econ772stf/gls.pdf


P.S.: please note the [math] prepended to the subject line above.
Since this list is shared for all commons components, we ask that
subject lines of posts to this list start with a reference to the
component that the question or comment refers to.

---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@commons.apache.org
For additional commands, e-mail: user-help@commons.apache.org


Mime
View raw message