Frank,
The algorithms are different. The default algorithm for
SAS ARIMA is a conditional least squares with backcasting the
starting values. Stata employs ML estimation for ARIMA.
Also, SAS can employ automatic prewhitening whereas Stata
does not, if it is set up differently.
Regards,
Bob Yaffee
Robert A. Yaffee, Ph.D.
Research Professor
Shirley M. Ehrenkranz
School of Social Work
New York University
home address:
Apt 19-W
2100 Linwood Ave.
Fort Lee, NJ
07024-3171
Phone: 201-242-3824
Fax: 201-242-3825
[email protected]
----- Original Message -----
From: Frank Zhang <[email protected]>
Date: Thursday, April 27, 2006 2:50 pm
Subject: st: transfer function model and regression: EQUIVALENT???
> Dear Statalisters,
>
> I have a question about transfer function model with
> ARIMA and REGRESSION precedures.
> If the model specification is the same, in my view,
> both of the following procedures should have the same
> results.
>
> proc reg data=one;
> model y= y1 x ;
> run;
> proc arima data=one;
> identify var=y crosscorr=x;
> estimate p=1 input=(x );
> run;
>
> But it turned out that the above two procedures give
> quite different results. CONFUSING!
> Can anybody tell me why? THANK YOU!
>
> The data and code in SAS are as follows:
>
> ---------------------------
> data one;
> input x y;
> y1=lag(y);
> cards;
> 1 -0.109 53.8
> 2 0 53.6
> 3 0.178 53.5
> 4 0.339 53.5
> 5 0.373 53.4
> 6 0.441 53.1
> 7 0.461 52.7
> 8 0.348 52.4
> 9 0.127 52.2
> 10 -0.18 52
> 11 -0.588 52
> 12 -1.055 52.4
> 13 -1.421 53
> 14 -1.52 54
> 15 -1.302 54.9
> 16 -0.814 56
> 17 -0.475 56.8
> 18 -0.193 56.8
> 19 0.088 56.4
> 20 0.435 55.7
> 21 0.771 55
> 22 0.866 54.3
> 23 0.875 53.2
> 24 0.891 52.3
> 25 0.987 51.6
> 26 1.263 51.2
> 27 1.775 50.8
> 28 1.976 50.5
> 29 1.934 50
> 30 1.866 49.2
> 31 1.832 48.4
> 32 1.767 47.9
> 33 1.608 47.6
> 34 1.265 47.5
> 35 0.79 47.5
> 36 0.36 47.6
> 37 0.115 48.1
> 38 0.088 49
> 39 0.331 50
> 40 0.645 51.1
> 41 0.96 51.8
> 42 1.409 51.9
> 43 2.67 51.7
> 44 2.834 51.2
> 45 2.812 50
> 46 2.483 48.3
> 47 1.929 47
> 48 1.485 45.8
> 49 1.214 45.6
> 50 1.239 46
> ;
> run;
> proc arima data=one;
> identify var=y crosscorr=x;
> estimate p=1 input=(x );
> run;
> proc reg;
> model y= y1 x ;
> run;
>
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam? Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
