Date: Tue, 21 Nov 2006 09:55:37 +0200
Reply-To: ATAKAN YALCIN <atyalcin@KU.EDU.TR>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: ATAKAN YALCIN <atyalcin@KU.EDU.TR>
Subject: Re: tests of heteroscedasticity with Proc Reg and Proc Model
Content-Type: text/plain; charset=US-ASCII
Thanks David. One final question about the panel data setup. Ideally, I
would like to run the panel data regression acounting for
heteroscedasticity, autocorrelation and crosscorrelation. The problem is
that I have 100 crosssections and only 35 timeseries which won't let me
correct for crosscorrelation (at least using PROC TSCSREG: when I use
the Parks method the phi matrix is singular). Any suggestions on this
one? One solution that comes to mind is to reduce the number of
crosssections (which is at my discretion)... however, I'd rather work
with 100 crosssections....
Assistant Professor of Finance
College of Administrative Sciences and Economics
Sariyer, Istanbul 34450, Turkey
>>> David L Cassell <davidlcassell@MSN.COM> 11/21/2006 8:18:08 AM >>>
>One can test for the presence of heteroskedasticity in many ways. Two
>that are commonly used in SAS are the SPEC option in Proc Reg, and
>White option in Proc Model. Now, I am aware that White's test in the
>MODEL procedure is different than White's test in the REG procedure
>requested by the SPEC option. The SPEC option produces the test from
>Theorem 2 on page 823 of White (1980). The WHITE option, on the other
>hand, produces the statistic from Corollary 1 on page 825 of White
>However, when I run these two tests, I get completely different
>results. One accepts the null hypothesis of homoscedasticity as the
>other is rejecting. Is there a way to settle the score? Or,
>alternatively, is one test better than the other under some special
First, let me congratulate you on doing your research on this.
So few people do. Sincec you have done your work, I am
encouraged to give you some help.
Unfortunately, all I can do at this point is tell you that you are
on the right track. The two tests are not identical. They have
slightly different sets of assumptions. If you validate each of
the underlying assumptions, then you may find that one of the
tests has the desired assumptions while the other has a violation
so its results are not usable. This would tell you which test
to trust. Of course, it is possible that neither test has all of
its assumptions met, and both results may be unreliable.
>I have a panel data set with 100 crosssections and 35 time series. I
>would like to estimate a linear model in this panel data setup. I
>to test for heteroscedasticty and the presence of autocorrelation
>making any attempts to correct for them. The first thing I do is to
>consider 35 different crosssections, and test the null hypothesis of
>homoscedasticity using the SPEC option in Proc Reg, or the White
>in proc model. These two tests give quite different results...and I
>can't proceed any further before resolving this matter and figuring
>what to do...
>I am looking forward to your suggestions. Thanks in advance.
Let me point you at PROC TSCSREG instead. You have SAS/ETS,
since you are using PROC MODEL. PROC TSCSREG, and PROC PANEL
if you have it, will automatically do the work of fitting this type of
problem without your having to put a lot of work in. Then you can try
the same thing in PROC REG and see whether your parameter estimates
are even in the right ballpark.
David L. Cassell
3115 NW Norwood Pl.
Corvallis OR 97330
View Athlete's Collections with Live Search