Date: Tue, 22 May 2007 04:38:44 -0400
Reply-To: Laprose Terry <laprose4769@YAHOO.CO.UK>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Laprose Terry <laprose4769@YAHOO.CO.UK>
Subject: Re: proc tscsreg
Content-Type: text/plain; charset=ISO-8859-1
Thanks for the advice; very helpful indeed.
Regarding the dummy: I have basically created a dummy which has 4
categories based on the values of an independent variable (W)which is a
continuous variable. The reason i did this is similar to the ones' when
running proc reg by dummy. I would basically like to examine the
coefficient of the variable of interest W, when the dummy is 1, 2, 3, 4.
The SAS manual notes that you can use by variables when using proc
tscsreg.So basically i run the proc tscsreg by dummy.
so here is a sample of my code
proc sort data=mydata;
proc tscsreg data =mydata;
model y= W C B P/ FIXTWO;
id ID YEAR;
WHERE DUMMY=1 IF W=
DUMMY=2 IF W=
DUMMY=3 IF W=
DUMMY=4 IF W=
On Mon, 21 May 2007 23:05:23 -0700, David L Cassell <davidlcassell@MSN.COM>
>>I am using this procedure using a panel dataset. Two questions:
>>1) Is it possible to compute heteroscedascticity adjusted (white) errors
>>using this procedure?
>No. The underlying assumptions of White's test are really restrictive,
>and the assumptions needed to do adjsutements for heteroskedasticity
>in panel data are complex. Really complex. Feel free to delve into
>something like Wayne Fuller's papers on this.
>But I don't see that you *need* this sort of adjusted error. Are you
>fitting the correct model? What kind of heteroskedasticity are you
>Can you transform your Y's or one of your regressors to remove this?
>Can you change your model to remove this? There are a lot of things
>you can do before you resort to the stuff that is *hard* to do.
>>2)Using the same dataset i run this proc by dummy. However, because the
>>dataset is unbalanced the number of observations is severely restricted.
>>Can anybody suggest whether another procedure such as proc mixed would be
>What do you mean 'by dummy'? Do you mean that you created scads
>of dummy variables instead of one regressor? This could really impact
>the covariance estimation process, you know. Why do you want to
>David L. Cassell
>3115 NW Norwood Pl.
>Corvallis OR 97330
>More photos, more messages, more storage�get 2GB with Windows Live