st: RE: RE: aweights and cook's distance

["Boim, Jason L. Mr. CIV OSD/DoDEA" <Jason.Boim@hq.dodea.edu>]

Sure. I am running a number of regressions on student test data, one for
each grade, against mostly demographic covariates. The data is
aggregated by school and grade, and I am weighting on the number of
students within each school-grade. There is a large amount of
variability, both within some of the covariates as well as in the number
of students (in some grades, the largest school is more than 200 times
that of the smallest). However, there are also some areas of certain
demographics that are rather sparse, which occasionally leads to
outliers and influential points. The regression coefficients will later
be used for an index, so I would rather have those influential points
removed to get a more accurate estimate for the majority of the data.
However, I'm not familiar with how (if?) cook's distance works with
aweights in regression, and wanted to check if it retains the same or
similar interpretation.

As far as output goes, what did you have in mind?

Best,

Jason

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of William
Buchanan
Sent: Friday, August 03, 2012 11:04 AM
To: statalist@hsphsun2.harvard.edu
Subject: st: RE: aweights and cook's distance

Hi Jason,

It might also be helpful to provide an example of your model and/or more
detailed information regarding the weights to get a better response from
the list.  If you can provide some output (of course without any
information that would be related to any students/faculty/staff in
DoDEA) that could also yield some more useful information for soliciting
feedback.

HTH,
Billy 

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Boim, Jason
L.
Mr. CIV OSD/DoDEA
Sent: Friday, August 03, 2012 7:58 AM
To: statalist@hsphsun2.harvard.edu
Subject: st: aweights and cook's distance

Hi All,

I am working with a weighted regression using analytic weights and
wanted to use cook's distance to check for influential points. It
occurred to me that the points with more weight would inherently be more
influential and as such was wondering if cook's distance could still be
used.

Thank you for your consideration,

Jason


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/