[prev in list] [next in list] [prev in thread] [next in thread] 

List:       sas-l
Subject:    Time-series Econometrics: The Newey-West Estimator
From:       "Manolakos, Peter" <pmanolakos () ISO-NE ! COM>
Date:       2003-08-29 20:51:45
[Download RAW message or body]

Hello,

This question is partly theoretical and partly SAS-related.  First, it is
well-known that under heteroscedasticity
and autocorrelation the classical estimator of the covariance matrix is
incorrect.  There are numerous approaches to obtaining a consistent
estimator under these conditions.  One is to use the so-called Newey-West
estimator.

Now, I have gathered that the specification of this estimator is a function
of the order of autocorrelation.  Thus if
we have MA(1) errors, then the estimator takes a different form than under
MA(7) errors.  In Proc Model, one might specify:

proc model data=b;
        by month ;
      rt_mean =  b0 + b1*da_mean ;
      %ma(rt_mean, 1);
      fit rt_mean / kernel=(bart,2,0);
      parms b0, b1;
run;

And expect to obtain the appropriate standard errors.  But when I do this it
produces the classical estimates.  It
also does not help to specify the gmm option for the fit statement.  When I
do so, sometimes the estimates do not
converge.

Any thoughts?

Peter

----------
Peter T. Manolakos
Associate Market Analyst
ISO-NE, Inc.
Market Monitoring Group
Ph. (413) 540-4233
----------

[prev in list] [next in list] [prev in thread] [next in thread]