Date:         Fri, 8 Aug 2008 09:26:10 -0400
Reply-To:     "Keintz, H. Mark" <mkeintz@WHARTON.UPENN.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Keintz, H. Mark" <mkeintz@WHARTON.UPENN.EDU>
Subject:      Re: What does "Interactivity disabled with BY processing" mean?
Comments: To: "du820531@GMAIL.COM" <du820531@GMAIL.COM>
In-Reply-To:  <aa7ab90e-fd39-44a1-93e3-5c4d729beedb@2g2000hsn.googlegroups.com>
Content-Type: text/plain; charset="us-ascii"

DU820531 asked about a message from PROC REG:



> -----Original Message-----
> From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of
> du820531@GMAIL.COM
> Sent: Thursday, August 07, 2008 7:55 PM
> To: SAS-L@LISTSERV.UGA.EDU
> Subject: What does "Interactivity disabled with BY processing" mean?
>
> I am doing a rolling regression on a dataset with the following
> structure:
> gvkey rollid year RET15 E_MV d_E_MV
>
> The output comes out but with a .log file that tells me following. I
> wonder whether the results are affected in some way and are they still
> usable. Thank you
>
> 25         ods listing close;
> 26         ods output ParameterEstimates=proj.valparest
> 26       ! FitStatistics=proj.valfit;
> 27         proc reg data=valdata;
> NOTE: The query requires remerging summary statistics back with the
>       original data.
> 28         model RET15= E_MV d_E_MV;
> 29         by gvkey rollid;
> 30         run;
>
> NOTE: Interactivity disabled with BY processing.
> NOTE: PROCEDURE REG used (Total process time):
>       real time           52.00 seconds
>       cpu time            38.34 seconds
>
> NOTE: The data set PROJ.VALFIT has 82875 observations and 10
> variables.
> NOTE: The data set PROJ.VALPAREST has 82875 observations and 10
> variables.


DU820531:

Rest assured that this note does NOT refer to a
statistical property of your models, but to how
PROC REG can be used.

Ordinarily (without a BY statement) in PROC REG
you can "interactively" issue a sequence of
MODEL .. RUN code blocks.  I.e., after seeing
the result of the first MODEL .. RUN, you can
spontaneously add/remove variables and issue a
second MODEL .. RUN.  This works, because you
are still "interacting" with PROC REG (you need
a QUIT statement to exit the proc)

The benefit of this approach is that the second
MODEL .. RUN block doesn't generate a rereading
of the data.  Instead it uses the SSCP matrix
produced by the first model, which is much faster.
(Note, if your first model uses only a subset of
vars you want in the SSCP matrix, then you need
to insert a VAR statement prior to the first
model statement.

BUT ... when using a BY statement, the ability
to issue the second model statement is not
supported, presumably because PROC REG is not
able to maintain an SSCP matrix for each BY level.

Regards,
Mark