From: Hazem Daouk <hdaouk@indiana.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Exogenous variables in VAR
Date: Fri, 1 May 1998 00:26:31 -0500 (EST)
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
X-Mailer: Mercury MTS (Bindery) v1.30

Dear rats users,

I am new to VARs. I have a model with endogenous and exogenous variables.
The exogenous variables have many lags. I know that to describe the effect
of an endogenous variable on another endogenous variable we use impulse
response functions. However, I was wondering if there is a way to describe
the effect of an EXOGENOUS variable on an endogenous variable (I cannot
just look at the coefficients since I have many lags of the exogenous
variable and the exogenous variable might be in many equations at a time).
The procedures described in the RATS manual only deal with the response
to shocks from endogenous variables. Can anybody give some advice (or
program) on how to deal with this issue? 

Thanks in advance,

Hazem Daouk
Department of Finance
Kelley School of Business
Indiana University
Bloomington, IN 47406



---------- End of message ----------

From: "Saravanan Balaraman" <estima@estima.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: Exogenous variables in VAR
Date: Fri, 1 May 1998 11:55:17 +0000
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
X-mailer: Pegasus Mail for Windows (v2.54) (via Mercury MTS (Bindery) v1.30)

> From:          Hazem Daouk <hdaouk@indiana.edu>
> To:            "RATS Discussion List" <RATS-L@efs.mq.edu.au>
> Subject:       Exogenous variables in VAR
> Date:          Fri, 1 May 1998 00:26:31 -0500 (EST)
> Reply-to:      "RATS Discussion List" <RATS-L@efs.mq.edu.au>

> Dear rats users,
> 
> I am new to VARs. I have a model with endogenous and exogenous variables.
> The exogenous variables have many lags. I know that to describe the effect
> of an endogenous variable on another endogenous variable we use impulse
> response functions. However, I was wondering if there is a way to describe
> the effect of an EXOGENOUS variable on an endogenous variable (I cannot
> just look at the coefficients since I have many lags of the exogenous
> variable and the exogenous variable might be in many equations at a time).
> The procedures described in the RATS manual only deal with the response
> to shocks from endogenous variables. Can anybody give some advice (or
> program) on how to deal with this issue? 
> 
> Thanks in advance,
> 
> Hazem Daouk
> Department of Finance
> Kelley School of Business
> Indiana University
> Bloomington, IN 47406
> 
> 
> 
> 
Dear Hazem,

			The right way of approaching this issue is to endogenise the 
exogenous variables i.e. express the dynamics of the endogenous 
variables in terms of equations and include the in the system. The 
section Near-VAR's in the RATS 4.0 Users Manual on page 8-2 explains 
this.  Also, the section Using Partial VAR's on page 8-25.


Sincerely,

Saravanan
Estima
--
+-----------------------------+-----------------------------------------+
| Estima                      |                                         |
| P.O. Box 1818               |  Voice: (847) 864-8772                  |
| Evanston, IL 60204-1818     |  Fax:   (847) 864-6221                  | 
| U.S.A                       |  BBS:   (847) 864-8816                  |
| e-mail: estima@estima.com   |  CompuServe: 73140,2202                 |
|-----------------------------------------------------------------------|
| Web Site:  http://www.estima.com                                      |
| RATS Internet Mailing List: New members can join by sending e-mail to |
|  MAISER@EFS.MQ.EDU.AU with the message: SUBSCRIBE  RATS-L             |
+-----------------------------------------------------------------------+

---------- End of message ----------

From: Shih-Jui  Wu <sw76@columbia.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re:unsubscribe
Date: Sat, 2 May 1998 16:34:01 -0400 (EDT)
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
X-Mailer: Mercury MTS (Bindery) v1.30

unsubscribe





---------- End of message ----------

From: Jae-Kwang Hwang <jhwang1@bcc.cba.ua.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: whiteness for residuals
Date: Sun, 03 May 1998 20:49:09 -0500
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
X-Mailer: Mozilla 4.04 [en] (Win95; I) (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Dear rats users,
I need to set optimal lag length in VAR estimation by using
general-to-specific modelling but I do not know to check the ressiduals
for whiteness. As you know, if the residuals proved to be non-white,
choose  a higher lag structure until they were whitened. Please send me
an answer how to check the residuals for whiteness in VAR.
Thank you so much for your help.


---------- End of message ----------

From: Simon van Norden <svn@alum.mit.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Internal Workings of LINREG
Date: Mon, 04 May 1998 00:05:29 -0400
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
Organization: Ecole des Hautes Etudes Commerciales
X-Mailer: Mozilla 4.03 [en] (Win95; I) (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit

Here's a puzzler I hope someone can help me solve.

I have two programs which run LINREG in a loop.
I thought that both programs *should* produce exactly the same results.
The first program changes the variables very slightly each time through
the loop.
The second program just changes the %CMOM matrix and then does
LINREG(CMOM).

In both programs, every pass through the loop has LINREG give the same
%CMOM, %BETA and %XX, %NOBS and %NDF.  However, the %RSS, %SEESQ and
%DURBIN differ across the two programs, as do the results of subsequent
hypothesis testing instructions.

Question: How could two otherwise-identical-looking LINREG statements
with the same %CMOM, %BETA and %XX give different %RSS, %SEESQ, etc.?
In particular, I'm wondering whether %RSS and %SEESQ are calculated by
going back to the original variables or whether they should be coming
straight from the %CMOM matrix.

I,d be grateful for any light anyone could shed on this.  I'd be happy
to post my code if anyone would like to see it.

Regards,
SvN
--
Simon van Norden        Professeur invité
simon.van-norden@hec.ca or svn@alum.mit.edu or http://www.hec.ca/~p280
(514)340-6781 or fax (514)340-5632 or cell (514)993-0466
Service de l'enseignement de la finance, Ecole des H.E.C.
3000 Cote-Sainte-Catherine, Montreal QC, CANADA   H3T 2A7




---------- End of message ----------

From: "Estima" <estima@estima.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: Internal Workings of LINREG
Date: Mon, 4 May 1998 14:30:28 -0600
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
X-mailer: Pegasus Mail for Win32 (v2.54) (via Mercury MTS (Bindery) v1.30)

> 
> Question: How could two otherwise-identical-looking LINREG statements
> with the same %CMOM, %BETA and %XX give different %RSS, %SEESQ, etc.?
> In particular, I'm wondering whether %RSS and %SEESQ are calculated by
> going back to the original variables or whether they should be coming
> straight from the %CMOM matrix.
> 

Yes, RATS uses the original data for those, which is why they change 
if the data changes (even if %CMOM stays the same). While it would be 
possible to compute some of these statistics (including RSS and 
SEESQ) from the CMOM matrix, RATS doesn't do it that way. I think the 
reason for this is that other statistics (e.g., Durbin-Watson) 
require the actual residuals anyway, so those are used for all the 
other summary stats as well.

I will try to make sure this is documented more clearly in the next 
edition of the manual. 

Sincerely,
Tom Maycock

------------------------------------------------------------
|   Estima                    |  Sales:   (800) 822-8038   |
|   P.O. Box 1818             |  Support: (847) 864-1910   |
|   Evanston, IL 60204-1818   |  Fax:     (847) 864-6221   |
|   USA                       |  estima@estima.com         |
|                             |  http://www.estima.com     |
------------------------------------------------------------

---------- End of message ----------

From: "Coleman, Mark" <MColeman@dlj.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: 'Compressing' %NA out of series
Date: Tue, 5 May 1998 15:44:09 -0400 
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.1960.3) (via Mercury MTS (Bindery) v1.30)
Content-Type: text/plain

Greetings:

I am working with a fairly large cross-sectional data set (about
n=40,000 individuals). For each individual I have about 40
characteristics. I am working with a standard probit model, so I've set
calendar from 1 to n as my baseline. I then read in a datafile which is
organized by obs, giving my 40 distinct series of size n. Thus, each
individual is coded to a specific "row"

I am looking for a way to "compress" or remove any %NA that appear in a
given series. Since each %NA correpsonds to a specific individual, I
effectively want to remove a subsample of the individuals from each of
the characteristics series, while naturally preserving each individual
with the correct row. 

For instance, one of my explanatory variables contains the year that the
individual entered the data set. I would like to be able to specify a
cut-off value for the year and then "remove" all of the individuals
("rows") from the sample and then store this abbreviated sample in a
RATS file. It's staightforward to indentify which rows to remove (I flag
them by converting all row values to %NA). Since the data is not
organized by year, however, the %NA appear in random points within
middle of the dataset. Thus, if I flag a cutoff year of 1990, then 40%
of my sample is flagged. I would like to take the remaining 60%, write
them to contiguous parts of the series, and then store the result over
the interval 1 to 0.60*n. To date, all I've done is apply the ORDER
command to the entire set of flagged series at once. This places all of
the %NA at the end of each series. then I can write the subsample to a
data file.

I was wondering if anyone had a better way to do this. The next round of
research will have a far larger n (n>500,000), and I'd like to avoid the
sorting overhead if possible.

Thanks

Mark


---------- End of message ----------

From: "Christopher F Baum" <baum@bc.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: 'Compressing' %NA out of series
Date: Tue, 05 May 1998 16:02:52 -0400
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
X-Mailer: Mulberry (MacOS) [1.3.3, s/n P020-300786-009] (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

This may be taken as a nonhelpful answer on this list, but the 'better way
to do this' is indubitably to use Stata (www.stata.com). Doing this sort of
cross-sectional, hierarchical, or panel data handling is something far
easier in Stata than in RATS, as is handling of 1+ million observation
datasets. On the other hand, I would not try to do heavy-duty time series
modeling in Stata. Sometimes different tools are best applied to different
problems. Both programs have their relative strengths.

I teach first-year PhD econometrics introducing students to both programs
for that very reason.

Kit Baum
Boston College

--On Tue, May 5, 1998 15:44 -0400 "Coleman, Mark" <MColeman@dlj.com> wrote: 

> Greetings:
> 
> I am working with a fairly large cross-sectional data set (about
> n=40,000 individuals). For each individual I have about 40
> characteristics. I am working with a standard probit model, so I've set
> calendar from 1 to n as my baseline. I then read in a datafile which is
> organized by obs, giving my 40 distinct series of size n. Thus, each
> individual is coded to a specific "row"
> 
> I am looking for a way to "compress" or remove any %NA that appear in a
> given series. Since each %NA correpsonds to a specific individual, I
> effectively want to remove a subsample of the individuals from each of
> the characteristics series, while naturally preserving each individual
> with the correct row. 
> 
> For instance, one of my explanatory variables contains the year that the
> individual entered the data set. I would like to be able to specify a
> cut-off value for the year and then "remove" all of the individuals
> ("rows") from the sample and then store this abbreviated sample in a
> RATS file. It's staightforward to indentify which rows to remove (I flag
> them by converting all row values to %NA). Since the data is not
> organized by year, however, the %NA appear in random points within
> middle of the dataset. Thus, if I flag a cutoff year of 1990, then 40%
> of my sample is flagged. I would like to take the remaining 60%, write
> them to contiguous parts of the series, and then store the result over
> the interval 1 to 0.60*n. To date, all I've done is apply the ORDER
> command to the entire set of flagged series at once. This places all of
> the %NA at the end of each series. then I can write the subsample to a
> data file.
> 
> I was wondering if anyone had a better way to do this. The next round of
> research will have a far larger n (n>500,000), and I'd like to avoid the
> sorting overhead if possible.
> 
> Thanks
> 
> Mark
> 



---------- End of message ----------

From: "Estima" <estima@estima.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: 'Compressing' %NA out of series
Date: Tue, 5 May 1998 16:30:21 -0600
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
X-mailer: Pegasus Mail for Win32 (v2.54) (via Mercury MTS (Bindery) v1.30)

> 
> I am working with a fairly large cross-sectional data set (about
> n=40,000 individuals). For each individual I have about 40
> characteristics. I am working with a standard probit model, so I've set
> calendar from 1 to n as my baseline. I then read in a datafile which is
> organized by obs, giving my 40 distinct series of size n. Thus, each
> individual is coded to a specific "row"
> 
> I am looking for a way to "compress" or remove any %NA that appear in a
> given series. Since each %NA correpsonds to a specific individual, I
> effectively want to remove a subsample of the individuals from each of
> the characteristics series, while naturally preserving each individual
> with the correct row. 
> 

Mark:

If I understand your setup, I think the SAMPLE command will work 
faster than the ORDER command for you. 

Say you have a series X for which the desired values have already 
been flagged as NA's. You can produce XSHORT (a compressed version of 
the series) by doing:

SAMPLE(SMPL=X) X / XSHORT

If you haven't already flagged the observations, you can just set a 
dummy variable according to the criteria in question. For example, 
suppose you have a series YEAR, and you want to eliminate all 
observations from X, Y, and Z for which YEAR is >1990. you could do:

SET DUMMY = YEAR<=1990
SAMPLE(SMPL=DUMMY) X / XSHORT
SAMPLE(SMPL=DUMMY) Y / YSHORT
SAMPLE(SMPL=DUMMY) Z / ZSHORT

If you have lots of series to compress, you could simplify this by 
storing the input series and the output (compressed) series in 
VECTORS or SERIES. For example, if the input series are stored in a 
vector of series called X, you could do:

SET DUMMY = YEAR<=1990
DECLARE VEC[SERIES] XSHORT(40)
DO i=1,40
 SAMPLE(SMPL=DUMMY) X(i) / XSHORT(i)
END 

I hope this is helpful.

Sincerely,
Tom Maycock
Estima


------------------------------------------------------------
|   Estima                    |  Sales:   (800) 822-8038   |
|   P.O. Box 1818             |  Support: (847) 864-1910   |
|   Evanston, IL 60204-1818   |  Fax:     (847) 864-6221   |
|   USA                       |  estima@estima.com         |
|                             |  http://www.estima.com     |
------------------------------------------------------------

---------- End of message ----------

From: mike@mier.po.my (Michael Yap)
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Sorry for the error
Date: Tue, 12 May 1998 14:49:27 +0800 (MYT)
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
X-Mailer: Mercury MTS (Bindery) v1.30

Dear All,

I am so sorry if you received an earlier e-mail meant for Estima.  It was
wrongly queued to the RATS Discussion List instead.  By the time I realised
this, it was too late.

Once again, my humble apologies.

Michael Yap

-----------------------------------------------------------
Michael Yap
Malaysian Institute of Economic Research
9th Floor, Block C, Bank Negara Malaysia
Jalan Kuching
P.O. Box 12160
50768 Kuala Lumpur
Malaysia.
Tel:603-292 6188; Fax:603-292 6163; E-mail:mike@mier.po.my
-----------------------------------------------------------


---------- End of message ----------

From: mike@mier.po.my (Michael Yap)
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Change of address
Date: Tue, 12 May 1998 14:49:15 +0800 (MYT)
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
X-Mailer: Mercury MTS (Bindery) v1.30

Dear Sir,

I am a registered user of WinRATS-32 (serial number WEJ479).  With effect
from 1 June 1998, my office will be moving to the address below:

        Michael Yap
        Malaysian Institute of Economic Research
        9th Floor, Menara Dayabumi
        Jalan Sultan Hishamuddin
        P.O. Box 12160
        50768 Kuala Lumpur
        Malaysia.

I shall be most appreciative if you could keep me on the mailing list of
the RATS news letter and any product upgrade information.

My e-mail address remains unchanged.

Thank you.

Best regards,
Michael Yap

P.S. By the way, what happened to the RATS news letter?

-----------------------------------------------------------
Michael Yap
Malaysian Institute of Economic Research
9th Floor, Block C, Bank Negara Malaysia
Jalan Kuching
P.O. Box 12160
50768 Kuala Lumpur
Malaysia.
Tel:603-292 6188; Fax:603-292 6163; E-mail:mike@mier.po.my
-----------------------------------------------------------


---------- End of message ----------

From: Felix.Maag@unisg.ch
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Newey/West estimator
Date: Tue, 12 May 1998 14:22:28 +0200
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
Mime-Version: 1.0
Content-type: multipart/mixed; 
X-Mailer: Mercury MTS (Bindery) v1.30

--0__=hsLhdlysElhK3Zli3MGb8tgpV72SoSkhMxqhA99VkWPVYZ4ZoTceSOfX
Content-type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-transfer-encoding: quoted-printable


Hi.

I have problems in implementing the Newey/West(1987) estimator.

The definition of  the Newey/West (1987) estimator is as follows:
(Hamilton p. 219 und p. 283)


(1)(Embedded image moved to file: pic06909.pcx)(Embedded image moved to=

file: pic12053.pcx)(Embedded image moved to file: pic20724.pcx)

Is it possible to implement this estimator with standard procedures in
Rats?
In the Rats manual 14-158 it is written that for  theta =3D 1 in the
expression

(2)  (Embedded image moved to file: pic32213.pcx)

together with (manual 14-157)

(3)  (Embedded image moved to file: pic30956.pcx)

should be the Newey/West(1987) estimator.

To implement this estimator I proceede as follows

linreg(robusterrors, lags=3D4, damp=3D1) urs1
# constant  uri5

Now the problem is that I don?t get the same results as in Eviews (lags=
 =3D
4) in which the
Newey/West (1987) (1) estimator is a standard procedure. I think that t=
he
difference
is because in RATS it sums over -L.......L in  (2)und (3) and not as
proposed by
Newy/West(1987) only from v =3D 1 to q in (1).

Is it possible to implement the proper Newey/West estimator (1) in Rats=
? If
yes, how?
Why is it written in the manual (14-158) that (2) should be the Newey/W=
est
(1987)
estimator for damp =3D1. As far as I see it thats not true (I hope that=
 I?m
wrong actually). But whats the
difference? I need the Newey/West estimator (1) to correct for
autocorrelation and heteroskedasticity.
Is that possible?


Thanks for help.

Felix
=

--0__=hsLhdlysElhK3Zli3MGb8tgpV72SoSkhMxqhA99VkWPVYZ4ZoTceSOfX
Content-type: application/octet-stream; 
	name="pic06909.pcx"
Content-Disposition: attachment; filename="pic06909.pcx"
Content-transfer-encoding: base64
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--0__=hsLhdlysElhK3Zli3MGb8tgpV72SoSkhMxqhA99VkWPVYZ4ZoTceSOfX
Content-type: application/octet-stream; 
	name="pic12053.pcx"
Content-Disposition: attachment; filename="pic12053.pcx"
Content-transfer-encoding: base64
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--0__=hsLhdlysElhK3Zli3MGb8tgpV72SoSkhMxqhA99VkWPVYZ4ZoTceSOfX
Content-type: application/octet-stream; 
	name="pic20724.pcx"
Content-Disposition: attachment; filename="pic20724.pcx"
Content-transfer-encoding: base64
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--0__=hsLhdlysElhK3Zli3MGb8tgpV72SoSkhMxqhA99VkWPVYZ4ZoTceSOfX
Content-type: application/octet-stream; 
	name="pic32213.pcx"
Content-Disposition: attachment; filename="pic32213.pcx"
Content-transfer-encoding: base64
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=

--0__=hsLhdlysElhK3Zli3MGb8tgpV72SoSkhMxqhA99VkWPVYZ4ZoTceSOfX
Content-type: application/octet-stream; 
	name="pic30956.pcx"
Content-Disposition: attachment; filename="pic30956.pcx"
Content-transfer-encoding: base64
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--0__=hsLhdlysElhK3Zli3MGb8tgpV72SoSkhMxqhA99VkWPVYZ4ZoTceSOfX--


---------- End of message ----------

From: "Frieder Knüpling" <knueplin@ruf.uni-freiburg.de>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: comment lines.
Date: Thu, 14 May 1998 13:08:37 +0200
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
X-Mailer: Mozilla 4.04 [en] (Win95; I) (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Dear RATS users,

is there a way to import coment lines for time series directly from
non-RATS-formats, especially from spreadsheats (.wks or .xls), apart
from cutting them in the spreadsheat and pasting them in RATSDATA (which
is inconvenient and has the danger of confusing them)?

Frieder

--
Frieder Knuepling
Albert-Ludwigs-Universitaet Freiburg
Institut fuer Allgemeine Wirtschaftsforschung
Abteilung Statistik und Oekonometrie
Belfortstr. 24
D-79098 Freiburg
Tel +49 761 / 203 - 2341
Fax +49 761 / 203 - 2340



---------- End of message ----------

From: "Estima" <estima@estima.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Estima e-mail address
Date: Thu, 14 May 1998 13:41:17 -0600
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
X-mailer: Pegasus Mail for Win32 (v2.54) (via Mercury MTS (Bindery) v1.30)

Dear RATS Users:

It appears that messages sent to our old "keynes" e-mail address 
(estima@keynes.acns.nwu.edu) is no longer being forwarded to us 
(we knew this would stop eventually, we just didn't know when). 

Unfortunately, mail sent to that address doesn't get bounced back to
the sender, but also never gets to us.

So, if you hadn't already done so, please be sure to update your 
records, e-mail address books, etc., to use our correct e-mail 
address:

estima@estima.com

I'm not sure when mail stopped being forwarded, but it may have been 
as much as several weeks ago. So, if you sent something to that 
address recently and haven't heard back from us, resend it to 
estima@estima.com

Thanks,
Tom Maycock
Estima

------------------------------------------------------------
|   Estima                    |  Sales:   (800) 822-8038   |
|   P.O. Box 1818             |  Support: (847) 864-1910   |
|   Evanston, IL 60204-1818   |  Fax:     (847) 864-6221   |
|   USA                       |  estima@estima.com         |
|                             |  http://www.estima.com     |
------------------------------------------------------------

---------- End of message ----------

From: WOO KAI YIN <hkwoo@hkabc.net>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: inquiry about simulation of Johansen's critical values
Date: Sat, 16 May 1998 01:22:15 +0800
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
X-Mailer: Mozilla 3.0 (Win95; I) (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Dear RATS users,

I wish to simulate Johansen's critical values for a VAR model with
intervention dummies(a constant with a number of structural breaks). I
know it depends on the sample size, the lag order and the number of
dummies. Do you know how Can I use RATS commands to perform such
simulation? (such simulation program can be written in GAUSS, can it be
written in RATS? ) 

Thank you!

Regards,

Kevin

---------- End of message ----------

From: Simon van Norden <svn@alum.mit.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: inquiry about simulation of Johansen's critical values
Date: Fri, 15 May 1998 13:54:03 -0400
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
Organization: Ecole des Hautes Etudes Commerciales
X-Mailer: Mozilla 4.03 [en] (Win95; I) (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit

Dear Woo;

The simulating part should be easy.  However, you will need some code
(preferably a procedure) which just returns the statistics in which you are
interested and does not require any user intervention (i.e. it has to run
in batch rather than interactive mode.)

Do you have such RATS or GAUSS code for the Johansen test?
Or is your problem that you are not sure how to simulate your model and do
the bookkeeping?

Simon

WOO KAI YIN wrote:

> Dear RATS users,
>
> I wish to simulate Johansen's critical values for a VAR model with
> intervention dummies(a constant with a number of structural breaks). I
> know it depends on the sample size, the lag order and the number of
> dummies. Do you know how Can I use RATS commands to perform such
> simulation? (such simulation program can be written in GAUSS, can it be
> written in RATS? )
>
> Thank you!
>
> Regards,
>
> Kevin



--
Simon van Norden        Professeur invité
simon.van-norden@hec.ca or svn@alum.mit.edu or http://www.hec.ca/~p280
(514)340-6781 or fax (514)340-5632 or cell (514)993-0466
Service de l'enseignement de la finance, Ecole des H.E.C.
3000 Cote-Sainte-Catherine, Montreal QC, CANADA   H3T 2A7



---------- End of message ----------

From: "David R.Tufte" <drtef@jazz.ucc.uno.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: RE: inquiry about simulation of Johansen's critical values
Date: Fri, 15 May 1998 17:44:14 -0500
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-version: 1.0
Content-type: multipart/mixed; boundary="---- =_NextPart_000_01BD804D.1F0FD3D0"
Content-transfer-encoding: 7BIT
X-Mailer: Mercury MTS (Bindery) v1.30


------ =_NextPart_000_01BD804D.1F0FD3D0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable

Your message was not specific enough. However, I think you should =
consider the program DisCo, which is available on Soren Johansen's web =
page. There is a link to this from the Estima home page.

David Tufte
Assistant Professor
Department of Economics and Finance
University of New Orleans
New Orleans, LA 70148

drtef@uno.edu

(504) 280-7094
(504) 280-6397 (fax)


-----Original Message-----
From:	WOO KAI YIN [SMTP:hkwoo@hkabc.net]
Sent:	Friday, May 15, 1998 12:22 PM
To:	RATS Discussion List
Subject:	inquiry about simulation of Johansen's critical values

Dear RATS users,

I wish to simulate Johansen's critical values for a VAR model with
intervention dummies(a constant with a number of structural breaks). I
know it depends on the sample size, the lag order and the number of
dummies. Do you know how Can I use RATS commands to perform such
simulation? (such simulation program can be written in GAUSS, can it be
written in RATS? )=20

Thank you!

Regards,

Kevin
------ =_NextPart_000_01BD804D.1F0FD3D0
Content-Type: application/ms-tnef
Content-Transfer-Encoding: base64
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------ =_NextPart_000_01BD804D.1F0FD3D0--


---------- End of message ----------

From: WOO KAI YIN <hkwoo@hkabc.net>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: inquiry about simulation of Johansen's critical values
Date: Sun, 17 May 1998 01:54:06 +0800
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
X-Mailer: Mozilla 3.0 (Win95; I) (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

David R.Tufte wrote:
> 
> Your message was not specific enough. However, I think you should consider the program DisCo, which is available on Soren Johansen's web page. There is a link to this from the Estima home page.
> 
> David Tufte
> Assistant Professor
> Department of Economics and Finance
> University of New Orleans
> New Orleans, LA 70148
> 
> drtef@uno.edu
> 
> (504) 280-7094
> (504) 280-6397 (fax)
> 
> -----Original Message-----
> From:   WOO KAI YIN [SMTP:hkwoo@hkabc.net]
> Sent:   Friday, May 15, 1998 12:22 PM
> To:     RATS Discussion List
> Subject:        inquiry about simulation of Johansen's critical values
> 
> Dear RATS users,
> 
> I wish to simulate Johansen's critical values for a VAR model with
> intervention dummies(a constant with a number of structural breaks). I
> know it depends on the sample size, the lag order and the number of
> dummies. Do you know how Can I use RATS commands to perform such
> simulation? (such simulation program can be written in GAUSS, can it be
> written in RATS? )
> 
> Thank you!
> 
> Regards,
> 
> Kevin
> 
>     ---------------------------------------------------------------
> 
>                 Part 1.2       Type: application/ms-tnef
>                            Encoding: base64
Dear Sir,

Thank you for your suggestion. But DisCo does not solve my bivariate VAR
model with so many dummies. 

Thank you!

Regards,

Kevin

---------- End of message ----------

From: YPORTILL@macrocon.com.pe (Yolanda Portilla)
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: help for E-garch
Date: Mon, 18 May 1998 21:46:57 +0000
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
X-mailer: Pegasus Mail for Windows (v2.52) (via Mercury MTS (Bindery) v1.30)
Organization: MACROCONSULT

I wanna know abuut T-GARCH AND E-GARCH in RATS. If exists procedures 
about this technics please explain me.


Thank you
Yolanda Portilla Sotomayor
Economist
Pontificia Universidad Catolica del Peru

---------- End of message ----------

From: "Cohen, Gerald (RSCH)" <Gerald_Cohen@ml.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Multicollinearity with Multinomial Logit
Date: Wed, 20 May 1998 11:03:38 -0400
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.1960.3) (via Mercury MTS (Bindery) v1.30)
Content-Type: text/plain

Does anybody have any suggestions how to deal with Multicollinearity
when estimating a model using Multinomial Logit.

Thank you in advance for your assistance.

Gerald D. Cohen
Vice President & Senior Economist
Merrill Lynch
Corporate Strategy and Research
World Financial Center
North Tower, 19th Floor
New York, NY 10281-1319
(212)449-0938
Fax: (212)449-8133
Gerald_Cohen@ml.com


---------- End of message ----------

From: "Estima" <estima@estima.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Hansen Stability Test, Historical Data
Date: Wed, 20 May 1998 11:07:48 -0600
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
X-mailer: Pegasus Mail for Win32 (v2.54) (via Mercury MTS (Bindery) v1.30)

Dear Folks:

We're working on a procedure that implements a parameter stability
test (Hansen, "Parameter Instability in Linear Models", J. of Policy
Modeling, 1992). 

In order to test the results against those in the paper, I've been
trying to get a hold of the historical GNP data put together by C.
Romer at Cal-Berkely (in '89, if I recall correctly). This is annual
US Real GNP dating back to 1889 (through 1987). 

If anyone has any tips as to where I might find this online or
elsewhere that would save me a trip to the library, I'd appreciate
it.

Thanks,
Tom Maycock
Estima


------------------------------------------------------------
|   Estima                    |  Sales:   (800) 822-8038   |
|   P.O. Box 1818             |  Support: (847) 864-1910   |
|   Evanston, IL 60204-1818   |  Fax:     (847) 864-6221   |
|   USA                       |  estima@estima.com         |
|                             |  http://www.estima.com     |
------------------------------------------------------------

---------- End of message ----------

From: Durwood Marshall <dmarshal@emerald.tufts.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: Multicollinearity with Multinomial Logit
Date: Wed, 20 May 1998 12:45:30 -0400 (EDT)
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-version: 1.0
Content-type: TEXT/PLAIN; charset=US-ASCII
X-Mailer: Mercury MTS (Bindery) v1.30

On Wed, 20 May 1998, Cohen, Gerald (RSCH) wrote:
> Does anybody have any suggestions how to deal with Multicollinearity
> when estimating a model using Multinomial Logit.

You might try subjecting the matrix of covariates(X) including the
intercept(if modeled) to some standard collinearity diagnostics.
Check: Belsley, Kuh and Welsch, Regression Diagnostics; Identifying
Influential Data and Sources of Collinearity, Wiley, ISBN 0-471-05856-4

Although the setting is linear regression there is still some benefit of 
checking the columns of data matrix(X) for possible dependencies by
running an auxillary regression.  There could be several and the
Variance-Decomposition matrix will provide insight into these.
Many packages offer something along these lines as an option to
regression.  SAS comes to mind for a preprogrammed solution, but you can
perform the calculations in RATS as well.   
 
Of course there are other approaches.  Standardize each column of X about 
it's mean to ensure approximate prior independence,
removing columns, shrinkage estimation and  ridge regression,
transformations and more.  A choice depends somewhat on the goals.
Hope this helps.

Durwood Marshall-Tufts University
Statistical and Research Computing Consulting
Tufts Computer and Communications Services
617.628.5000 x2180 : Fax 617.627.3667 : E-Mail: dmarshal@tufts.edu
-----------------------------------------------------------------
The real reason why aliens don't come forth:
  Life on earth was the result of an alien genetic experiment gone bad.
  They are embarrassed and willing to wait to see if everything
  turns out ok, or not....everyone makes mistakes!



---------- End of message ----------

From: Norman Morin <m1njm00@Frb.GOV>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: Hansen Stability Test, Historical Data 
Date: Fri, 22 May 1998 09:43:53 -0400
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
X-Mailer: exmh version 1.6.9.4 5/23/97 (via Mercury MTS (Bindery) v1.30)
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii


> We're working on a procedure that implements a parameter stability
> test (Hansen, "Parameter Instability in Linear Models", J. of Policy
> Modeling, 1992). 
> 
> In order to test the results against those in the paper, I've been
> trying to get a hold of the historical GNP data put together by C.
> Romer at Cal-Berkely (in '89, if I recall correctly). This is annual
> US Real GNP dating back to 1889 (through 1987). 
> 
> If anyone has any tips as to where I might find this online or
> elsewhere that would save me a trip to the library, I'd appreciate
> it.
> 

Tom, 

Does that test have a nonstandard/tabulated distribution?  If you cannot
get the dataset, you could always do a Monte Carlo with your proc
and see if the critical values are close to his.  I double-checked
a proc I wrote to do the Zivot-Andrews unit root test against a
level-shift/trend-break alternative w/ and unknown break date doing 
that.

Norm
-- 

                           Norman J. Morin 
                  nmorin@frb.gov or m1njm00@frb.gov         
          --------------------------------------------------
          Division of Research and Statistics * Mail Stop 82
           Board of Governors of the Federal Reserve System 
               Washington, D.C. 20551 * (202) 452-2476
          --------------------------------------------------



---------- End of message ----------

From: "Mahmoud H. Al-Osaimy" <ECF3006@KAAU.EDU.SA>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Rolling Regression
Date: Fri, 22 May 98 17:36:40 G+3
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
X-Mailer: Mercury MTS (Bindery) v1.30


Hi all :

What is the Rolling Regression? What is the advantage of it?.
Some references will be appreciated.

Regards,

Mahmoud____

---------- End of message ----------

From: "Christopher F Baum" <baum@bc.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: Hansen Stability Test, Historical Data 
Date: Fri, 22 May 1998 12:11:08 -0400
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
X-Mailer: Mulberry (MacOS) [1.3.3, s/n P020-300786-009] (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Bruce Hansen has posted the original data and GAUSS programs used on his web
page; I let Tom know of this by a private communication yesterday. 

http://www2.bc.edu/~hansenb

Kit Baum
Boston College

--On Fri, May 22, 1998 9:43 -0400 "Norman Morin" <m1njm00@Frb.GOV> wrote: 

> 
>> We're working on a procedure that implements a parameter stability
>> test (Hansen, "Parameter Instability in Linear Models", J. of Policy
>> Modeling, 1992). 
>> 
>> In order to test the results against those in the paper, I've been
>> trying to get a hold of the historical GNP data put together by C.
>> Romer at Cal-Berkely (in '89, if I recall correctly). This is annual
>> US Real GNP dating back to 1889 (through 1987). 
>> 
>> If anyone has any tips as to where I might find this online or
>> elsewhere that would save me a trip to the library, I'd appreciate
>> it.
>> 
> 
> Tom, 
> 
> Does that test have a nonstandard/tabulated distribution?  If you cannot
> get the dataset, you could always do a Monte Carlo with your proc
> and see if the critical values are close to his.  I double-checked
> a proc I wrote to do the Zivot-Andrews unit root test against a
> level-shift/trend-break alternative w/ and unknown break date doing 
> that.
> 
> Norm
> -- 
> 
>                            Norman J. Morin 
>                   nmorin@frb.gov or m1njm00@frb.gov         
>           --------------------------------------------------
>           Division of Research and Statistics * Mail Stop 82
>            Board of Governors of the Federal Reserve System 
>                Washington, D.C. 20551 * (202) 452-2476
>           --------------------------------------------------
> 
> 



---------- End of message ----------

From: "Wai Lee"<lee_wai@jpmorgan.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: RE: rolling regression
Date: Fri, 22 May 1998 13:08:45 -0400
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
Mime-Version: 1.0
Content-type: text/plain; charset=US-ASCII
X-Mailer: Mercury MTS (Bindery) v1.30


DO J = 1, 100
     COMPUTE START = 1959:12 + J
     COMPUTE END = START + 120
     SMPL START END
     LINREG(NOPRINT) Y
     # CONSTANT X

     * some book keeping instructions for results of regression etc.

END DO J

The above is a 10-year rolling regression.  You roll your sample period 1
month at a time, and run the same regression in rolling windows.  We
ususally keep track of how the regression results change over time, or to
make out-of-sample forecasts from there.

Hope it's useful.

W. Lee



---------- End of message ----------

From: Simon van Norden <svn@alum.mit.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: rolling regression
Date: Fri, 22 May 1998 13:36:55 -0400
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
Organization: Ecole des Hautes Etudes Commerciales
X-Mailer: Mozilla 4.03 [en] (Win95; I) (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit

I wrote some code a while back that put most everything you'd want to do into
a procedure (i.e. constant, growing or shrinking sample size, adding or
dropping observations, returning or graphing series of coefficient
estimates.)  It also did efficient computation (updated the %CMOM matrix
instead of recalculating it every time) for speed. I did a similar one for
chow tests, as I recall.

You can find these procs and others on my web page (below).

If there's demand, I could post the rollreg proc to the list.

--
Simon van Norden        Professeur invité
simon.van-norden@hec.ca or svn@alum.mit.edu or http://www.hec.ca/~p280
(514)340-6781 or fax (514)340-5632 or cell (514)993-0466
Service de l'enseignement de la finance, Ecole des H.E.C.
3000 Cote-Sainte-Catherine, Montreal QC, CANADA   H3T 2A7



---------- End of message ----------

From: "Cheesman, Fred" <fcheesman@ncsc.dni.us>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Prewhitening Time series for Cross correlation
Date: Fri, 22 May 98 14:26:11 EST
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
X-Mailer: Mercury MTS (Bindery) v1.30


     Is it possible to use RATS to "prewhiten" two time series before 
     calculating their cross-correlation function, as a preliminary step in 
     developing a transfer function model?  How might this be accomplished?
     Any recommendations for software to do such calculations (outside of 
     SAS)? Regards to my colleagues.
     
     Cheers!
     
     Fred Cheesman
     National Center for State Courts 

---------- End of message ----------

From: Simon van Norden <svn@alum.mit.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: rolling regressions
Date: Sun, 24 May 1998 12:41:42 -0400
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
Organization: Ecole des Hautes Etudes Commerciales
X-Mailer: Mozilla 4.03 [en] (Win95; I) (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit

I got a couple of requests for the code, so I'm posting it to the list.

--
Simon van Norden        Professeur invité
simon.van-norden@hec.ca or svn@alum.mit.edu or http://www.hec.ca/~p280
(514)340-6781 or fax (514)340-5632 or cell (514)993-0466
Service de l'enseignement de la finance, Ecole des H.E.C.
3000 Cote-Sainte-Catherine, Montreal QC, CANADA   H3T 2A7

 env noecho
*
* ROLLREG.SRC            by Simon van Norden           10-11-1992
* Copyright 1993, 1992, 1990 by Bank of Canada
* (modified for RATS v4 August 1993 by Jeff Gable)
* (Robust Error Bug Fix  June 1995 by Rob Vigfusson)
*
* This procedure does rolling OLS regressions in one of three modes.
* Given an intial range of observations, ADD adds observations
* (one at a time) to the end of the sample, DROP drops them from the
* beginning of the sample, and MOVE simultaneously adds and drops
* so that the number of observations in each regression is constant.
*
* The syntax is
* @ROLLREG  depvar  first last
* # list of series
* where
* depvar            is the name of the LHS variable
* list of series    lists the RHS variables for the regression
* first, last       (optional) maximum range for the regression
*                   defaults to maximum range for which all variables
*                   are defined
* Options:
* ADD=              starting with a regression from  first  to  ADD,
*                   adds one observations at a time until it does a
*                   regression from  first  to  last
* DROP=             starting with a regression from  first  to  last,
*                   drops one observations at a time until it does a
*                   regression from  DROP  to  last
* MOVE=             starting with a regression from first to
first+MOVE-1
*                   adds and drops one observations at a time until it
*                   does a regression from  last-move+1  to  last
*                   (i.e. there are MOVE observations in each
regression)
*
* Note that exactly one of these three options must be used.
* Also note that if you use dates with any of the above options, you
must
* enclose the dates in parentheses.
* e.g. @rollreg(drop=(89,4)) gnp
*      @rollreg(add=(71,4)) gnp
*      @rollreg(move=20) gnp
*
* ROBUST/[NOROBUST] uses the ROBUSTERRORS option to calculate the
*                   standard errors.
* LAGS = [0]        number of lags to use with the ROBUST option.
*                   This does **not** vary with the degrees of freedom.
* DAMP = [1.0]      shape of the spectral window to use with the
*                   ROBUST option.  Note that the default here is a
*                   Newey-West or Bartlett window while in RATS the
*                   default is a flat window (DAMP=0.0).
*
* [PRINT]/NOPRINT   prints the coefficient estimates
* GRAPH/[NOGRAPH]   graphs the test statistics.  At present, the graphs
can be
*                   converted to postscript format using the command:
*                   >>gsp2pst chow.gsp post-name<<
*      They can then be printed using the lpr command or viewed
*                   using the postscript viewer.    You must issue the
"open
*                   plot" and "close plot" commands from the calling
program.
* COPY/[NOCOPY]     writes the estimated coefficients to a file.
*                   You must have already opened the copy unit to this
*                   file
* FORMAT = [FREE]/BINARY/WKS/DIF/PRN/EXCEL/TROLL
*                   for use only with COPY; writes the data in that
*                   format.  Note that ORG=VAR.
*
procedure rollreg  y  first last
type series[real] y
type integer  first last
option integer ADD     0
option integer DROP    0
option integer MOVE    0
option switch robust   0
option real damp      1.0
option integer lags    0
option switch print    1
option switch graph    0
option switch copy     0
option choice format   1 free binary wks dif prn excel troll
local vector[integer] x
local integer fobs lobs numreg base cut k   fbeta lbeta stdbase
local vector[real] xtyt
local rectangular[real] xy
local series[real] m
dis; dis
*
* Define the system to be estimated and initialize
*
enter(varying,entries=numreg) x
inquire(regressorlist) fobs lobs
# y x
if (first .and. last)
    {;  compute fobs=first;   compute lobs=last;    }
dis 'Rolling Regression from' %datelabel(fobs) 'to' %datelabel(lobs) $
    'of ' %label(y) ' on '
dis(hold) '    '
do k=1,numreg
   dis(hold)  %label([series]x(k))  '  '
end do k
dis
make(transpose) xy fobs lobs
# x y
if (add==0)+(drop==0)+(move==0) < 2
    dis 'ROLLREG ERROR: Use only one of ADD, DROP and MOVE options.'
else if .not. (add .or. drop .or. move)
    dis 'ROLLREG ERROR: Must specify one of ADD, DROP and MOVE options.'

else if add
    {
*
* Adding observations to end
*
    dis 'using ADD=' %datelabel(add)
    scratch 2*numreg add lobs base
    cmoment fobs add
    # x y
    linreg(cmom,noprint,robusterrors=robust,damp=damp,lags=lags) y
    # x
    dofor m = base+1 to base+numreg
        compute m(add) = %beta(m-base)
    end dofor
 if robust == 0
     dofor m=base+1+numreg to base+2*numreg
    compute m(add) = sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
     end dofor
 else if robust == 1
    dofor m=base+1+numreg to base+2*numreg
      compute m(add) = sqrt(%xx(m-base-numreg,m-base-numreg))
      end dofor


    do cut=add+1,lobs
 cmoment(setup) fobs cut
 # x y
        overlay xy(1,cut-fobs+1) with xtyt(numreg+1)
        compute %cmom = %cmom + xtyt*tr(xtyt)
        linreg(cmom,noprint,robusterrors=robust,damp=damp,lags=lags) y
        # x
        dofor m=base+1 to base+numreg
            compute m(cut) = %beta(m-base)
        end dofor
 if robust == 0
         dofor m=base+1+numreg to base+2*numreg
             compute m(cut) =
sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
         end dofor
 else if robust == 1
  dofor m=base+1+numreg to base+2*numreg
              compute m(cut) = sqrt(%xx(m-base-numreg,m-base-numreg))
         end dofor

    end do cut
    compute fbeta = add; compute lbeta = lobs
    }
else if drop
    {
*
* Dropping observations from start
*
    dis 'using DROP=' %datelabel(drop)
    scratch 2*numreg fobs drop base
    cmoment fobs lobs
    # x y
    linreg(cmom,noprint,robusterrors=robust,damp=damp,lags=lags) y
    # x
    dofor m=base+1 to base+numreg
        compute m(fobs) = %beta(m-base)
    end dofor
    if robust == 0
     dofor m=base+1+numreg to base+2*numreg
         compute m(fobs) = sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)

     end dofor
    else if robust == 1
 dofor m=base+1+numreg to base+2*numreg
         compute m(fobs) = sqrt(%xx(m-base-numreg,m-base-numreg))
     end dofor


    do cut=fobs,drop-1
 cmoment(setup) cut+1 lobs
 # x y
        overlay xy(1,cut-fobs+1) with xtyt(numreg+1)
        compute %cmom = %cmom - xtyt*tr(xtyt)
        linreg(cmom,noprint,robusterrors=robust,damp=damp,lags=lags) y
        # x
        dofor m=base+1 to base+numreg
            compute m(cut+1) = %beta(m-base)
        end dofor
 if robust == 0
        dofor m=base+1+numreg to base+2*numreg
            comptue m(cut+1) =
sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
        end dofor
 else if robust == 1
 dofor m=base+1+numreg to base+2*numreg
            comptue m(cut+1) = sqrt(%xx(m-base-numreg,m-base-numreg))
        end dofor



    end do cut
    compute fbeta = fobs; compute lbeta = drop
    }
else if MOVE
    {
*
* Moving sample regression
*
    dis 'using MOVE=' ### move
    scratch 2*numreg fobs+move-1 lobs base
    cmoment fobs fobs+move-1
    # x y
    linreg(cmom,noprint,robusterrors=robust,damp=damp,lags=lags) y
    # x
    dofor m=base+1 to base+numreg
        compute m(fobs+move-1) = %beta(m-base)
    end dofor
    if robust == 0
     dofor m=base+1+numreg to base+2*numreg
        compute m(fobs+move-1) =
sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
     end dofor
    else if robust == 1
  dofor m=base+1+numreg to base+2*numreg
        compute m(fobs+move-1) = sqrt(%xx(m-base-numreg,m-base-numreg))
 end dofor

    do cut=fobs+move,lobs
 cmoment(setup) cut-move+1 cut
 # x y
        overlay xy(1,cut-fobs+1) with xtyt(numreg+1)
        compute %cmom = %cmom + xtyt*tr(xtyt)
        overlay xy(1,cut-move-fobs+1) with xtyt(numreg+1)
        compute %cmom = %cmom - xtyt*tr(xtyt)
        linreg(cmom,noprint,robusterrors=robust,damp=damp,lags=lags) y
        # x
        dofor m=base+1 to base+numreg
            compute m(cut) = %beta(m-base)
        end dofor
 if robust == 0
         dofor m=base+1+numreg to base+2*numreg
            compute m(cut) =
sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
         end dofor
 else if robust == 1
   dofor m=base+1+numreg to base+2*numreg
            compute m(cut) = sqrt(%xx(m-base-numreg,m-base-numreg))
          end dofor
    end do cut
    compute fbeta = fobs+move-1; compute lbeta = lobs
    }
*
* Now output the results
*
if (add==0)+(drop==0)+(move==0) == 2
{
do k = 1,numreg
    labels base+k base+k+numreg
#  %label([series]x(k)) %concat('se',%label([series]x(k)))
end do k
if print
    print(dates) fbeta lbeta base+1 to base+2*numreg
if copy
    copy(format=format) fbeta lbeta  base+1 to base+2*numreg
if graph
    {
    scratch 2 fbeta lbeta stdbase
    labels stdbase+1 stdbase+2
    # '95%upper' '95%lower'
    do k = base+1, base+numreg
        set stdbase+1 fbeta lbeta = k + (k+numreg)*1.96
        set stdbase+2 fbeta lbeta = k - (k+numreg)*1.96
        if graph
            {
            graph(dates,header='rollreg estimated betas') 3
            # k fbeta lbeta
            # stdbase+1 fbeta lbeta
            # stdbase+2 fbeta lbeta
            }
    end do k
    }
dis 'Coefficients stored in series' ### base+1 'to' ### base+numreg
'and'
dis 'standard errors in series' ### base+numreg+1 'to' ### base+2*numreg

}
*
end proc rollreg
env echo

 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
*
 *  ROLLREG.SRC
 *  Copyright 1993 by the Bank of Canada
*
 *  Programmed for RATS by Robert Amano, Jeff Gable and Simon van Norden
*
 *  In exchange for access to these procedures, users are requested to
*
 *  acknowledge the use of the Bank of Canada RATS procedures in
*
 *  all published work.
*
 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
*




---------- End of message ----------

From: <W.Killeen@London.sbic.co.za> (by way of Rob Trevor)
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: RE: Rolling Regression
Date: Mon, 25 May 1998 15:06:53 +1000
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
X-Mailer: Mercury MTS (Bindery) v1.30


Rolling regression means implementing a regression model on success windows
or subsamples through a larger sample. The aim is to  assess parameter
constancy over time / identify structural breaks ; chow type analysis. This
sort of analysis is particularly useful for looking at the sample behavior
of long term parameters in principal components analysis (see CATS manual )
 ----------

Hi all :

What is the Rolling Regression? What is the advantage of it?.
Some references will be appreciated.

Regards,

Mahmoud____



---------- End of message ----------

From: YPORTILL@macrocon.com.pe (Yolanda Portilla)
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: a question about maxima verosimilitud
Date: Tue, 26 May 1998 05:00:48 +0000
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
X-mailer: Pegasus Mail for Windows (v2.52) (via Mercury MTS (Bindery) v1.30)
Organization: MACROCONSULT

I want to know about the posibility to put restrictions in the 
parameters in maximum likelihood estimation.
For example, I want that a<2.4.

I will thankful for your answers.


Thank you
Yolanda Portilla Sotomayor
Economist
Pontificia Universidad Catolica del Peru

---------- End of message ----------

From: "Christopher F Baum" <baum@bc.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: a question about maxima verosimilitud
Date: Wed, 27 May 1998 12:32:45 -0400
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
X-Mailer: Mulberry (MacOS) [1.3.3, s/n P020-300786-009] (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

A very effective method (that, alas, does not involve RATS) is using GAUSS'
Constrained Maximum Likelihood package (www.aptech.com), which allows for
linear equality constraints, linear inequality constraints, bounds
constraints, and nonlinear inequality constraints (e.g. the smallest
eigenvalue of the VCV > 1.0e-5). A very expensive solution, but I don't
think there's a good substitute. 
(That's what differentiates it from Windows, for which there is a very good
substitute).

Think Different!

Kit Baum
Boston College

--On Tue, May 26, 1998 5:00 +0000 "Yolanda Portilla"
<YPORTILL@macrocon.com.pe> wrote: 

> I want to know about the posibility to put restrictions in the 
> parameters in maximum likelihood estimation.
> For example, I want that a<2.4.
> 
> I will thankful for your answers.
> 
> 
> Thank you
> Yolanda Portilla Sotomayor
> Economist
> Pontificia Universidad Catolica del Peru



---------- End of message ----------

From: "Frieder Knüpling" <knueplin@ruf.uni-freiburg.de>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: a question about maxima verosimilitud
Date: Wed, 27 May 1998 19:32:40 +0200
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
X-Mailer: Mozilla 4.04 [en] (Win95; I) (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

What about replacing the parameter a by something like
     2.4-exp(b)
and using b as a free parameter instead? (That's cheap.)

Frieder

--
Frieder Knuepling
Albert-Ludwigs-Universitaet Freiburg
Institut fuer Allgemeine Wirtschaftsforschung
Abteilung Statistik und Oekonometrie
Belfortstr. 24
D-79098 Freiburg
Tel +49 761 / 203 - 2341
Fax +49 761 / 203 - 2340



---------- End of message ----------

From: "Skip Kreger" <regress@gte.net>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: rolling regressions, related question
Date: Wed, 27 May 1998 22:40:06 -0500
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
Content-Type: text/plain;
X-Mailer: Microsoft Outlook Express 4.71.1712.3 (via Mercury MTS (Bindery) v1.30)
Content-Transfer-Encoding: quoted-printable

Does anyone have any suggestions for appropriate estimation techniques if=
,
based on the rolling regressions, we find significant parameter instabili=
ty?

From my limited readings on the topic, one possibility would be to estima=
te
the equation with time-varying parameters using the Kalman Filter.  I
understand the math of a rudamentary time-varying parameter model (one
assuming a random walk in the parameter), but I haven't been able to figu=
re
out how to do the estimation or how to interpret the results.  Anybody tr=
ied
this, or have a good source?

Skip Krueger
PhD student
Political Science Department
University of North Texas


-----Original Message-----
From: Simon van Norden <svn@alum.mit.edu>
To: RATS Discussion List <RATS-L@efs.mq.edu.au>
Date: Sunday, May 24, 1998 11:55 AM
Subject: Re: rolling regressions


>I got a couple of requests for the code, so I'm posting it to the list.
>
>--
>Simon van Norden        Professeur invit=E9
>simon.van-norden@hec.ca or svn@alum.mit.edu or http://www.hec.ca/~p280
>(514)340-6781 or fax (514)340-5632 or cell (514)993-0466
>Service de l'enseignement de la finance, Ecole des H.E.C.
>3000 Cote-Sainte-Catherine, Montreal QC, CANADA   H3T 2A7
>
> env noecho
>*
>* ROLLREG.SRC            by Simon van Norden           10-11-1992
>* Copyright 1993, 1992, 1990 by Bank of Canada
>* (modified for RATS v4 August 1993 by Jeff Gable)
>* (Robust Error Bug Fix  June 1995 by Rob Vigfusson)
>*
>* This procedure does rolling OLS regressions in one of three modes.
>* Given an intial range of observations, ADD adds observations
>* (one at a time) to the end of the sample, DROP drops them from the
>* beginning of the sample, and MOVE simultaneously adds and drops
>* so that the number of observations in each regression is constant.
>*
>* The syntax is
>* @ROLLREG  depvar  first last
>* # list of series
>* where
>* depvar            is the name of the LHS variable
>* list of series    lists the RHS variables for the regression
>* first, last       (optional) maximum range for the regression
>*                   defaults to maximum range for which all variables
>*                   are defined
>* Options:
>* ADD=3D              starting with a regression from  first  to  ADD,
>*                   adds one observations at a time until it does a
>*                   regression from  first  to  last
>* DROP=3D             starting with a regression from  first  to  last,
>*                   drops one observations at a time until it does a
>*                   regression from  DROP  to  last
>* MOVE=3D             starting with a regression from first to
>first+MOVE-1
>*                   adds and drops one observations at a time until it
>*                   does a regression from  last-move+1  to  last
>*                   (i.e. there are MOVE observations in each
>regression)
>*
>* Note that exactly one of these three options must be used.
>* Also note that if you use dates with any of the above options, you
>must
>* enclose the dates in parentheses.
>* e.g. @rollreg(drop=3D(89,4)) gnp
>*      @rollreg(add=3D(71,4)) gnp
>*      @rollreg(move=3D20) gnp
>*
>* ROBUST/[NOROBUST] uses the ROBUSTERRORS option to calculate the
>*                   standard errors.
>* LAGS =3D [0]        number of lags to use with the ROBUST option.
>*                   This does **not** vary with the degrees of freedom.
>* DAMP =3D [1.0]      shape of the spectral window to use with the
>*                   ROBUST option.  Note that the default here is a
>*                   Newey-West or Bartlett window while in RATS the
>*                   default is a flat window (DAMP=3D0.0).
>*
>* [PRINT]/NOPRINT   prints the coefficient estimates
>* GRAPH/[NOGRAPH]   graphs the test statistics.  At present, the graphs
>can be
>*                   converted to postscript format using the command:
>*                   >>gsp2pst chow.gsp post-name<<
>*      They can then be printed using the lpr command or viewed
>*                   using the postscript viewer.    You must issue the
>"open
>*                   plot" and "close plot" commands from the calling
>program.
>* COPY/[NOCOPY]     writes the estimated coefficients to a file.
>*                   You must have already opened the copy unit to this
>*                   file
>* FORMAT =3D [FREE]/BINARY/WKS/DIF/PRN/EXCEL/TROLL
>*                   for use only with COPY; writes the data in that
>*                   format.  Note that ORG=3DVAR.
>*
>procedure rollreg  y  first last
>type series[real] y
>type integer  first last
>option integer ADD     0
>option integer DROP    0
>option integer MOVE    0
>option switch robust   0
>option real damp      1.0
>option integer lags    0
>option switch print    1
>option switch graph    0
>option switch copy     0
>option choice format   1 free binary wks dif prn excel troll
>local vector[integer] x
>local integer fobs lobs numreg base cut k   fbeta lbeta stdbase
>local vector[real] xtyt
>local rectangular[real] xy
>local series[real] m
>dis; dis
>*
>* Define the system to be estimated and initialize
>*
>enter(varying,entries=3Dnumreg) x
>inquire(regressorlist) fobs lobs
># y x
>if (first .and. last)
>    {;  compute fobs=3Dfirst;   compute lobs=3Dlast;    }
>dis 'Rolling Regression from' %datelabel(fobs) 'to' %datelabel(lobs) $
>    'of ' %label(y) ' on '
>dis(hold) '    '
>do k=3D1,numreg
>   dis(hold)  %label([series]x(k))  '  '
>end do k
>dis
>make(transpose) xy fobs lobs
># x y
>if (add=3D=3D0)+(drop=3D=3D0)+(move=3D=3D0) < 2
>    dis 'ROLLREG ERROR: Use only one of ADD, DROP and MOVE options.'
>else if .not. (add .or. drop .or. move)
>    dis 'ROLLREG ERROR: Must specify one of ADD, DROP and MOVE options.'
>
>else if add
>    {
>*
>* Adding observations to end
>*
>    dis 'using ADD=3D' %datelabel(add)
>    scratch 2*numreg add lobs base
>    cmoment fobs add
>    # x y
>    linreg(cmom,noprint,robusterrors=3Drobust,damp=3Ddamp,lags=3Dlags) y
>    # x
>    dofor m =3D base+1 to base+numreg
>        compute m(add) =3D %beta(m-base)
>    end dofor
> if robust =3D=3D 0
>     dofor m=3Dbase+1+numreg to base+2*numreg
>    compute m(add) =3D sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>     end dofor
> else if robust =3D=3D 1
>    dofor m=3Dbase+1+numreg to base+2*numreg
>      compute m(add) =3D sqrt(%xx(m-base-numreg,m-base-numreg))
>      end dofor
>
>
>    do cut=3Dadd+1,lobs
> cmoment(setup) fobs cut
> # x y
>        overlay xy(1,cut-fobs+1) with xtyt(numreg+1)
>        compute %cmom =3D %cmom + xtyt*tr(xtyt)
>        linreg(cmom,noprint,robusterrors=3Drobust,damp=3Ddamp,lags=3Dlag=
s) y
>        # x
>        dofor m=3Dbase+1 to base+numreg
>            compute m(cut) =3D %beta(m-base)
>        end dofor
> if robust =3D=3D 0
>         dofor m=3Dbase+1+numreg to base+2*numreg
>             compute m(cut) =3D
>sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>         end dofor
> else if robust =3D=3D 1
>  dofor m=3Dbase+1+numreg to base+2*numreg
>              compute m(cut) =3D sqrt(%xx(m-base-numreg,m-base-numreg))
>         end dofor
>
>    end do cut
>    compute fbeta =3D add; compute lbeta =3D lobs
>    }
>else if drop
>    {
>*
>* Dropping observations from start
>*
>    dis 'using DROP=3D' %datelabel(drop)
>    scratch 2*numreg fobs drop base
>    cmoment fobs lobs
>    # x y
>    linreg(cmom,noprint,robusterrors=3Drobust,damp=3Ddamp,lags=3Dlags) y
>    # x
>    dofor m=3Dbase+1 to base+numreg
>        compute m(fobs) =3D %beta(m-base)
>    end dofor
>    if robust =3D=3D 0
>     dofor m=3Dbase+1+numreg to base+2*numreg
>         compute m(fobs) =3D sqrt(%xx(m-base-numreg,m-base-numreg)*%sees=
q)
>
>     end dofor
>    else if robust =3D=3D 1
> dofor m=3Dbase+1+numreg to base+2*numreg
>         compute m(fobs) =3D sqrt(%xx(m-base-numreg,m-base-numreg))
>     end dofor
>
>
>    do cut=3Dfobs,drop-1
> cmoment(setup) cut+1 lobs
> # x y
>        overlay xy(1,cut-fobs+1) with xtyt(numreg+1)
>        compute %cmom =3D %cmom - xtyt*tr(xtyt)
>        linreg(cmom,noprint,robusterrors=3Drobust,damp=3Ddamp,lags=3Dlag=
s) y
>        # x
>        dofor m=3Dbase+1 to base+numreg
>            compute m(cut+1) =3D %beta(m-base)
>        end dofor
> if robust =3D=3D 0
>        dofor m=3Dbase+1+numreg to base+2*numreg
>            comptue m(cut+1) =3D
>sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>        end dofor
> else if robust =3D=3D 1
> dofor m=3Dbase+1+numreg to base+2*numreg
>            comptue m(cut+1) =3D sqrt(%xx(m-base-numreg,m-base-numreg))
>        end dofor
>
>
>
>    end do cut
>    compute fbeta =3D fobs; compute lbeta =3D drop
>    }
>else if MOVE
>    {
>*
>* Moving sample regression
>*
>    dis 'using MOVE=3D' ### move
>    scratch 2*numreg fobs+move-1 lobs base
>    cmoment fobs fobs+move-1
>    # x y
>    linreg(cmom,noprint,robusterrors=3Drobust,damp=3Ddamp,lags=3Dlags) y
>    # x
>    dofor m=3Dbase+1 to base+numreg
>        compute m(fobs+move-1) =3D %beta(m-base)
>    end dofor
>    if robust =3D=3D 0
>     dofor m=3Dbase+1+numreg to base+2*numreg
>        compute m(fobs+move-1) =3D
>sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>     end dofor
>    else if robust =3D=3D 1
>  dofor m=3Dbase+1+numreg to base+2*numreg
>        compute m(fobs+move-1) =3D sqrt(%xx(m-base-numreg,m-base-numreg)=
)
> end dofor
>
>    do cut=3Dfobs+move,lobs
> cmoment(setup) cut-move+1 cut
> # x y
>        overlay xy(1,cut-fobs+1) with xtyt(numreg+1)
>        compute %cmom =3D %cmom + xtyt*tr(xtyt)
>        overlay xy(1,cut-move-fobs+1) with xtyt(numreg+1)
>        compute %cmom =3D %cmom - xtyt*tr(xtyt)
>        linreg(cmom,noprint,robusterrors=3Drobust,damp=3Ddamp,lags=3Dlag=
s) y
>        # x
>        dofor m=3Dbase+1 to base+numreg
>            compute m(cut) =3D %beta(m-base)
>        end dofor
> if robust =3D=3D 0
>         dofor m=3Dbase+1+numreg to base+2*numreg
>            compute m(cut) =3D
>sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>         end dofor
> else if robust =3D=3D 1
>   dofor m=3Dbase+1+numreg to base+2*numreg
>            compute m(cut) =3D sqrt(%xx(m-base-numreg,m-base-numreg))
>          end dofor
>    end do cut
>    compute fbeta =3D fobs+move-1; compute lbeta =3D lobs
>    }
>*
>* Now output the results
>*
>if (add=3D=3D0)+(drop=3D=3D0)+(move=3D=3D0) =3D=3D 2
>{
>do k =3D 1,numreg
>    labels base+k base+k+numreg
>#  %label([series]x(k)) %concat('se',%label([series]x(k)))
>end do k
>if print
>    print(dates) fbeta lbeta base+1 to base+2*numreg
>if copy
>    copy(format=3Dformat) fbeta lbeta  base+1 to base+2*numreg
>if graph
>    {
>    scratch 2 fbeta lbeta stdbase
>    labels stdbase+1 stdbase+2
>    # '95%upper' '95%lower'
>    do k =3D base+1, base+numreg
>        set stdbase+1 fbeta lbeta =3D k + (k+numreg)*1.96
>        set stdbase+2 fbeta lbeta =3D k - (k+numreg)*1.96
>        if graph
>            {
>            graph(dates,header=3D'rollreg estimated betas') 3
>            # k fbeta lbeta
>            # stdbase+1 fbeta lbeta
>            # stdbase+2 fbeta lbeta
>            }
>    end do k
>    }
>dis 'Coefficients stored in series' ### base+1 'to' ### base+numreg
>'and'
>dis 'standard errors in series' ### base+numreg+1 'to' ### base+2*numreg
>
>}
>*
>end proc rollreg
>env echo
>
> * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
>*
> *  ROLLREG.SRC
> *  Copyright 1993 by the Bank of Canada
>*
> *  Programmed for RATS by Robert Amano, Jeff Gable and Simon van Norden
>*
> *  In exchange for access to these procedures, users are requested to
>*
> *  acknowledge the use of the Bank of Canada RATS procedures in
>*
> *  all published work.
>*
> * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
>*
>
>
>


---------- End of message ----------

From: "Skip Kreger" <regress@gte.net>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: rolling regressions, related question
Date: Wed, 27 May 1998 22:40:06 -0500
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
Content-Type: text/plain;
X-Mailer: Microsoft Outlook Express 4.71.1712.3 (via Mercury MTS (Bindery) v1.30
)
Content-Transfer-Encoding: quoted-printable

Does anyone have any suggestions for appropriate estimation techniques if=
,
based on the rolling regressions, we find significant parameter instabili=
ty?

From my limited readings on the topic, one possibility would be to estima=
te
the equation with time-varying parameters using the Kalman Filter.  I
understand the math of a rudamentary time-varying parameter model (one
assuming a random walk in the parameter), but I haven't been able to figu=
re
out how to do the estimation or how to interpret the results.  Anybody tr=
ied
this, or have a good source?

Skip Krueger
PhD student
Political Science Department
University of North Texas


-----Original Message-----
From: Simon van Norden <svn@alum.mit.edu>
To: RATS Discussion List <RATS-L@efs.mq.edu.au>
Date: Sunday, May 24, 1998 11:55 AM
Subject: Re: rolling regressions


>I got a couple of requests for the code, so I'm posting it to the list.
>
>--
>Simon van Norden        Professeur invit=E9
>simon.van-norden@hec.ca or svn@alum.mit.edu or http://www.hec.ca/~p280
>(514)340-6781 or fax (514)340-5632 or cell (514)993-0466
>Service de l'enseignement de la finance, Ecole des H.E.C.
>3000 Cote-Sainte-Catherine, Montreal QC, CANADA   H3T 2A7
>
> env noecho
>*
>* ROLLREG.SRC            by Simon van Norden           10-11-1992
>* Copyright 1993, 1992, 1990 by Bank of Canada
>* (modified for RATS v4 August 1993 by Jeff Gable)
>* (Robust Error Bug Fix  June 1995 by Rob Vigfusson)
>*
>* This procedure does rolling OLS regressions in one of three modes.
>* Given an intial range of observations, ADD adds observations
>* (one at a time) to the end of the sample, DROP drops them from the
>* beginning of the sample, and MOVE simultaneously adds and drops
>* so that the number of observations in each regression is constant.
>*
>* The syntax is
>* @ROLLREG  depvar  first last
>* # list of series
>* where
>* depvar            is the name of the LHS variable
>* list of series    lists the RHS variables for the regression
>* first, last       (optional) maximum range for the regression
>*                   defaults to maximum range for which all variables
>*                   are defined
>* Options:
>* ADD=3D              starting with a regression from  first  to  ADD,
>*                   adds one observations at a time until it does a
>*                   regression from  first  to  last
>* DROP=3D             starting with a regression from  first  to  last,
>*                   drops one observations at a time until it does a
>*                   regression from  DROP  to  last
>* MOVE=3D             starting with a regression from first to
>first+MOVE-1
>*                   adds and drops one observations at a time until it
>*                   does a regression from  last-move+1  to  last
>*                   (i.e. there are MOVE observations in each
>regression)
>*
>* Note that exactly one of these three options must be used.
>* Also note that if you use dates with any of the above options, you
>must
>* enclose the dates in parentheses.
>* e.g. @rollreg(drop=3D(89,4)) gnp
>*      @rollreg(add=3D(71,4)) gnp
>*      @rollreg(move=3D20) gnp
>*
>* ROBUST/[NOROBUST] uses the ROBUSTERRORS option to calculate the
>*                   standard errors.
>* LAGS =3D [0]        number of lags to use with the ROBUST option.
>*                   This does **not** vary with the degrees of freedom.
>* DAMP =3D [1.0]      shape of the spectral window to use with the
>*                   ROBUST option.  Note that the default here is a
>*                   Newey-West or Bartlett window while in RATS the
>*                   default is a flat window (DAMP=3D0.0).
>*
>* [PRINT]/NOPRINT   prints the coefficient estimates
>* GRAPH/[NOGRAPH]   graphs the test statistics.  At present, the graphs
>can be
>*                   converted to postscript format using the command:
>*                   >>gsp2pst chow.gsp post-name<<
>*      They can then be printed using the lpr command or viewed
>*                   using the postscript viewer.    You must issue the
>"open
>*                   plot" and "close plot" commands from the calling
>program.
>* COPY/[NOCOPY]     writes the estimated coefficients to a file.
>*                   You must have already opened the copy unit to this
>*                   file
>* FORMAT =3D [FREE]/BINARY/WKS/DIF/PRN/EXCEL/TROLL
>*                   for use only with COPY; writes the data in that
>*                   format.  Note that ORG=3DVAR.
>*
>procedure rollreg  y  first last
>type series[real] y
>type integer  first last
>option integer ADD     0
>option integer DROP    0
>option integer MOVE    0
>option switch robust   0
>option real damp      1.0
>option integer lags    0
>option switch print    1
>option switch graph    0
>option switch copy     0
>option choice format   1 free binary wks dif prn excel troll
>local vector[integer] x
>local integer fobs lobs numreg base cut k   fbeta lbeta stdbase
>local vector[real] xtyt
>local rectangular[real] xy
>local series[real] m
>dis; dis
>*
>* Define the system to be estimated and initialize
>*
>enter(varying,entries=3Dnumreg) x
>inquire(regressorlist) fobs lobs
># y x
>if (first .and. last)
>    {;  compute fobs=3Dfirst;   compute lobs=3Dlast;    }
>dis 'Rolling Regression from' %datelabel(fobs) 'to' %datelabel(lobs) $
>    'of ' %label(y) ' on '
>dis(hold) '    '
>do k=3D1,numreg
>   dis(hold)  %label([series]x(k))  '  '
>end do k
>dis
>make(transpose) xy fobs lobs
># x y
>if (add=3D=3D0)+(drop=3D=3D0)+(move=3D=3D0) < 2
>    dis 'ROLLREG ERROR: Use only one of ADD, DROP and MOVE options.'
>else if .not. (add .or. drop .or. move)
>    dis 'ROLLREG ERROR: Must specify one of ADD, DROP and MOVE options.'
>
>else if add
>    {
>*
>* Adding observations to end
>*
>    dis 'using ADD=3D' %datelabel(add)
>    scratch 2*numreg add lobs base
>    cmoment fobs add
>    # x y
>    linreg(cmom,noprint,robusterrors=3Drobust,damp=3Ddamp,lags=3Dlags) y
>    # x
>    dofor m =3D base+1 to base+numreg
>        compute m(add) =3D %beta(m-base)
>    end dofor
> if robust =3D=3D 0
>     dofor m=3Dbase+1+numreg to base+2*numreg
>    compute m(add) =3D sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>     end dofor
> else if robust =3D=3D 1
>    dofor m=3Dbase+1+numreg to base+2*numreg
>      compute m(add) =3D sqrt(%xx(m-base-numreg,m-base-numreg))
>      end dofor
>
>
>    do cut=3Dadd+1,lobs
> cmoment(setup) fobs cut
> # x y
>        overlay xy(1,cut-fobs+1) with xtyt(numreg+1)
>        compute %cmom =3D %cmom + xtyt*tr(xtyt)
>        linreg(cmom,noprint,robusterrors=3Drobust,damp=3Ddamp,lags=3Dlag=
s) y
>        # x
>        dofor m=3Dbase+1 to base+numreg
>            compute m(cut) =3D %beta(m-base)
>        end dofor
> if robust =3D=3D 0
>         dofor m=3Dbase+1+numreg to base+2*numreg
>             compute m(cut) =3D
>sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>         end dofor
> else if robust =3D=3D 1
>  dofor m=3Dbase+1+numreg to base+2*numreg
>              compute m(cut) =3D sqrt(%xx(m-base-numreg,m-base-numreg))
>         end dofor
>
>    end do cut
>    compute fbeta =3D add; compute lbeta =3D lobs
>    }
>else if drop
>    {
>*
>* Dropping observations from start
>*
>    dis 'using DROP=3D' %datelabel(drop)
>    scratch 2*numreg fobs drop base
>    cmoment fobs lobs
>    # x y
>    linreg(cmom,noprint,robusterrors=3Drobust,damp=3Ddamp,lags=3Dlags) y
>    # x
>    dofor m=3Dbase+1 to base+numreg
>        compute m(fobs) =3D %beta(m-base)
>    end dofor
>    if robust =3D=3D 0
>     dofor m=3Dbase+1+numreg to base+2*numreg
>         compute m(fobs) =3D sqrt(%xx(m-base-numreg,m-base-numreg)*%sees=
q)
>
>     end dofor
>    else if robust =3D=3D 1
> dofor m=3Dbase+1+numreg to base+2*numreg
>         compute m(fobs) =3D sqrt(%xx(m-base-numreg,m-base-numreg))
>     end dofor
>
>
>    do cut=3Dfobs,drop-1
> cmoment(setup) cut+1 lobs
> # x y
>        overlay xy(1,cut-fobs+1) with xtyt(numreg+1)
>        compute %cmom =3D %cmom - xtyt*tr(xtyt)
>        linreg(cmom,noprint,robusterrors=3Drobust,damp=3Ddamp,lags=3Dlag=
s) y
>        # x
>        dofor m=3Dbase+1 to base+numreg
>            compute m(cut+1) =3D %beta(m-base)
>        end dofor
> if robust =3D=3D 0
>        dofor m=3Dbase+1+numreg to base+2*numreg
>            comptue m(cut+1) =3D
>sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>        end dofor
> else if robust =3D=3D 1
> dofor m=3Dbase+1+numreg to base+2*numreg
>            comptue m(cut+1) =3D sqrt(%xx(m-base-numreg,m-base-numreg))
>        end dofor
>
>
>
>    end do cut
>    compute fbeta =3D fobs; compute lbeta =3D drop
>    }
>else if MOVE
>    {
>*
>* Moving sample regression
>*
>    dis 'using MOVE=3D' ### move
>    scratch 2*numreg fobs+move-1 lobs base
>    cmoment fobs fobs+move-1
>    # x y
>    linreg(cmom,noprint,robusterrors=3Drobust,damp=3Ddamp,lags=3Dlags) y
>    # x
>    dofor m=3Dbase+1 to base+numreg
>        compute m(fobs+move-1) =3D %beta(m-base)
>    end dofor
>    if robust =3D=3D 0
>     dofor m=3Dbase+1+numreg to base+2*numreg
>        compute m(fobs+move-1) =3D
>sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>     end dofor
>    else if robust =3D=3D 1
>  dofor m=3Dbase+1+numreg to base+2*numreg
>        compute m(fobs+move-1) =3D sqrt(%xx(m-base-numreg,m-base-numreg)=
)
> end dofor
>
>    do cut=3Dfobs+move,lobs
> cmoment(setup) cut-move+1 cut
> # x y
>        overlay xy(1,cut-fobs+1) with xtyt(numreg+1)
>        compute %cmom =3D %cmom + xtyt*tr(xtyt)
>        overlay xy(1,cut-move-fobs+1) with xtyt(numreg+1)
>        compute %cmom =3D %cmom - xtyt*tr(xtyt)
>        linreg(cmom,noprint,robusterrors=3Drobust,damp=3Ddamp,lags=3Dlag=
s) y
>        # x
>        dofor m=3Dbase+1 to base+numreg
>            compute m(cut) =3D %beta(m-base)
>        end dofor
> if robust =3D=3D 0
>         dofor m=3Dbase+1+numreg to base+2*numreg
>            compute m(cut) =3D
>sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>         end dofor
> else if robust =3D=3D 1
>   dofor m=3Dbase+1+numreg to base+2*numreg
>            compute m(cut) =3D sqrt(%xx(m-base-numreg,m-base-numreg))
>          end dofor
>    end do cut
>    compute fbeta =3D fobs+move-1; compute lbeta =3D lobs
>    }
>*
>* Now output the results
>*
>if (add=3D=3D0)+(drop=3D=3D0)+(move=3D=3D0) =3D=3D 2
>{
>do k =3D 1,numreg
>    labels base+k base+k+numreg
>#  %label([series]x(k)) %concat('se',%label([series]x(k)))
>end do k
>if print
>    print(dates) fbeta lbeta base+1 to base+2*numreg
>if copy
>    copy(format=3Dformat) fbeta lbeta  base+1 to base+2*numreg
>if graph
>    {
>    scratch 2 fbeta lbeta stdbase
>    labels stdbase+1 stdbase+2
>    # '95%upper' '95%lower'
>    do k =3D base+1, base+numreg
>        set stdbase+1 fbeta lbeta =3D k + (k+numreg)*1.96
>        set stdbase+2 fbeta lbeta =3D k - (k+numreg)*1.96
>        if graph
>            {
>            graph(dates,header=3D'rollreg estimated betas') 3
>            # k fbeta lbeta
>            # stdbase+1 fbeta lbeta
>            # stdbase+2 fbeta lbeta
>            }
>    end do k
>    }
>dis 'Coefficients stored in series' ### base+1 'to' ### base+numreg
>'and'
>dis 'standard errors in series' ### base+numreg+1 'to' ### base+2*numreg
>
>}
>*
>end proc rollreg
>env echo
>
> * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
>*
> *  ROLLREG.SRC
> *  Copyright 1993 by the Bank of Canada
>*
> *  Programmed for RATS by Robert Amano, Jeff Gable and Simon van Norden
>*
> *  In exchange for access to these procedures, users are requested to
>*
> *  acknowledge the use of the Bank of Canada RATS procedures in
>*
> *  all published work.
>*
> * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
>*
>
>
>


---------- End of message ----------

From: "Wai Lee" <lee_wai@jpmorgan.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: rolling regressions, related question
Date: Thu, 28 May 1998 08:42:48 -0400
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
Mime-Version: 1.0
Content-type: text/plain; charset=us-ascii
X-Mailer: Mercury MTS (Bindery) v1.30

Estima sent me a sample program (incomplete and nontested) two years ago to
do Kalman filtering with, say, AR(1) parameters instead of random walk.  I
modified it somehow to estimate stochastic volatilities models.  I guess a
much improved version of it should be available from Estima now.  Check
their homepage.

An alternative to Kalman filtering is Bayesian dynamic linear models, which
are related to, but more general than Kalman filtering.  This class of
models updates the parameters as new observations are collected.  Of
course, it is Bayesian updating.  Procedures in S-plus have been developed
by Jeff Harrison of U. of Warwick, and are made available to others.
Useful references include, West and Harrison, "Bayesian Forecasting and
Dynamic Models," 1997, 2nd Ed., Springer; and Pole, West, and Harrison,
"Applied Bayesian Forecasting and Time Series Analysis," 1994, Springer.  I
don't think similar source files in RATS exist.

Hope these help.

W. Lee






"Skip Kreger" <regress @ gte.net> on 05/27/98 11:40:06 PM

Please respond to "RATS Discussion List" <RATS-L@efs.mq.edu.au>

To:   "RATS Discussion List" <RATS-L @ efs.mq.edu.au>
cc:   (Wai Lee)
Subject:  rolling regressions, related question





Does anyone have any suggestions for appropriate estimation techniques if,
based on the rolling regressions, we find significant parameter
instability?

From my limited readings on the topic, one possibility would be to estimate
the equation with time-varying parameters using the Kalman Filter.  I
understand the math of a rudamentary time-varying parameter model (one
assuming a random walk in the parameter), but I haven't been able to figure
out how to do the estimation or how to interpret the results.  Anybody
tried
this, or have a good source?

Skip Krueger
PhD student
Political Science Department
University of North Texas


-----Original Message-----
From: Simon van Norden <svn@alum.mit.edu>
To: RATS Discussion List <RATS-L@efs.mq.edu.au>
Date: Sunday, May 24, 1998 11:55 AM
Subject: Re: rolling regressions


>I got a couple of requests for the code, so I'm posting it to the list.
>
>--
>Simon van Norden        Professeur invit?
>simon.van-norden@hec.ca or svn@alum.mit.edu or http://www.hec.ca/~p280
>(514)340-6781 or fax (514)340-5632 or cell (514)993-0466
>Service de l'enseignement de la finance, Ecole des H.E.C.
>3000 Cote-Sainte-Catherine, Montreal QC, CANADA   H3T 2A7
>
> env noecho
>*
>* ROLLREG.SRC            by Simon van Norden           10-11-1992
>* Copyright 1993, 1992, 1990 by Bank of Canada
>* (modified for RATS v4 August 1993 by Jeff Gable)
>* (Robust Error Bug Fix  June 1995 by Rob Vigfusson)
>*
>* This procedure does rolling OLS regressions in one of three modes.
>* Given an intial range of observations, ADD adds observations
>* (one at a time) to the end of the sample, DROP drops them from the
>* beginning of the sample, and MOVE simultaneously adds and drops
>* so that the number of observations in each regression is constant.
>*
>* The syntax is
>* @ROLLREG  depvar  first last
>* # list of series
>* where
>* depvar            is the name of the LHS variable
>* list of series    lists the RHS variables for the regression
>* first, last       (optional) maximum range for the regression
>*                   defaults to maximum range for which all variables
>*                   are defined
>* Options:
>* ADD=              starting with a regression from  first  to  ADD,
>*                   adds one observations at a time until it does a
>*                   regression from  first  to  last
>* DROP=             starting with a regression from  first  to  last,
>*                   drops one observations at a time until it does a
>*                   regression from  DROP  to  last
>* MOVE=             starting with a regression from first to
>first+MOVE-1
>*                   adds and drops one observations at a time until it
>*                   does a regression from  last-move+1  to  last
>*                   (i.e. there are MOVE observations in each
>regression)
>*
>* Note that exactly one of these three options must be used.
>* Also note that if you use dates with any of the above options, you
>must
>* enclose the dates in parentheses.
>* e.g. @rollreg(drop=(89,4)) gnp
>*      @rollreg(add=(71,4)) gnp
>*      @rollreg(move=20) gnp
>*
>* ROBUST/[NOROBUST] uses the ROBUSTERRORS option to calculate the
>*                   standard errors.
>* LAGS = [0]        number of lags to use with the ROBUST option.
>*                   This does **not** vary with the degrees of freedom.
>* DAMP = [1.0]      shape of the spectral window to use with the
>*                   ROBUST option.  Note that the default here is a
>*                   Newey-West or Bartlett window while in RATS the
>*                   default is a flat window (DAMP=0.0).
>*
>* [PRINT]/NOPRINT   prints the coefficient estimates
>* GRAPH/[NOGRAPH]   graphs the test statistics.  At present, the graphs
>can be
>*                   converted to postscript format using the command:
>*                   >>gsp2pst chow.gsp post-name<<
>*      They can then be printed using the lpr command or viewed
>*                   using the postscript viewer.    You must issue the
>"open
>*                   plot" and "close plot" commands from the calling
>program.
>* COPY/[NOCOPY]     writes the estimated coefficients to a file.
>*                   You must have already opened the copy unit to this
>*                   file
>* FORMAT = [FREE]/BINARY/WKS/DIF/PRN/EXCEL/TROLL
>*                   for use only with COPY; writes the data in that
>*                   format.  Note that ORG=VAR.
>*
>procedure rollreg  y  first last
>type series[real] y
>type integer  first last
>option integer ADD     0
>option integer DROP    0
>option integer MOVE    0
>option switch robust   0
>option real damp      1.0
>option integer lags    0
>option switch print    1
>option switch graph    0
>option switch copy     0
>option choice format   1 free binary wks dif prn excel troll
>local vector[integer] x
>local integer fobs lobs numreg base cut k   fbeta lbeta stdbase
>local vector[real] xtyt
>local rectangular[real] xy
>local series[real] m
>dis; dis
>*
>* Define the system to be estimated and initialize
>*
>enter(varying,entries=numreg) x
>inquire(regressorlist) fobs lobs
># y x
>if (first .and. last)
>    {;  compute fobs=first;   compute lobs=last;    }
>dis 'Rolling Regression from' %datelabel(fobs) 'to' %datelabel(lobs) $
>    'of ' %label(y) ' on '
>dis(hold) '    '
>do k=1,numreg
>   dis(hold)  %label([series]x(k))  '  '
>end do k
>dis
>make(transpose) xy fobs lobs
># x y
>if (add==0)+(drop==0)+(move==0) < 2
>    dis 'ROLLREG ERROR: Use only one of ADD, DROP and MOVE options.'
>else if .not. (add .or. drop .or. move)
>    dis 'ROLLREG ERROR: Must specify one of ADD, DROP and MOVE options.'
>
>else if add
>    {
>*
>* Adding observations to end
>*
>    dis 'using ADD=' %datelabel(add)
>    scratch 2*numreg add lobs base
>    cmoment fobs add
>    # x y
>    linreg(cmom,noprint,robusterrors=robust,damp=damp,lags=lags) y
>    # x
>    dofor m = base+1 to base+numreg
>        compute m(add) = %beta(m-base)
>    end dofor
> if robust == 0
>     dofor m=base+1+numreg to base+2*numreg
>    compute m(add) = sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>     end dofor
> else if robust == 1
>    dofor m=base+1+numreg to base+2*numreg
>      compute m(add) = sqrt(%xx(m-base-numreg,m-base-numreg))
>      end dofor
>
>
>    do cut=add+1,lobs
> cmoment(setup) fobs cut
> # x y
>        overlay xy(1,cut-fobs+1) with xtyt(numreg+1)
>        compute %cmom = %cmom + xtyt*tr(xtyt)
>        linreg(cmom,noprint,robusterrors=robust,damp=damp,lags=lags) y
>        # x
>        dofor m=base+1 to base+numreg
>            compute m(cut) = %beta(m-base)
>        end dofor
> if robust == 0
>         dofor m=base+1+numreg to base+2*numreg
>             compute m(cut) =
>sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>         end dofor
> else if robust == 1
>  dofor m=base+1+numreg to base+2*numreg
>              compute m(cut) = sqrt(%xx(m-base-numreg,m-base-numreg))
>         end dofor
>
>    end do cut
>    compute fbeta = add; compute lbeta = lobs
>    }
>else if drop
>    {
>*
>* Dropping observations from start
>*
>    dis 'using DROP=' %datelabel(drop)
>    scratch 2*numreg fobs drop base
>    cmoment fobs lobs
>    # x y
>    linreg(cmom,noprint,robusterrors=robust,damp=damp,lags=lags) y
>    # x
>    dofor m=base+1 to base+numreg
>        compute m(fobs) = %beta(m-base)
>    end dofor
>    if robust == 0
>     dofor m=base+1+numreg to base+2*numreg
>         compute m(fobs) = sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>
>     end dofor
>    else if robust == 1
> dofor m=base+1+numreg to base+2*numreg
>         compute m(fobs) = sqrt(%xx(m-base-numreg,m-base-numreg))
>     end dofor
>
>
>    do cut=fobs,drop-1
> cmoment(setup) cut+1 lobs
> # x y
>        overlay xy(1,cut-fobs+1) with xtyt(numreg+1)
>        compute %cmom = %cmom - xtyt*tr(xtyt)
>        linreg(cmom,noprint,robusterrors=robust,damp=damp,lags=lags) y
>        # x
>        dofor m=base+1 to base+numreg
>            compute m(cut+1) = %beta(m-base)
>        end dofor
> if robust == 0
>        dofor m=base+1+numreg to base+2*numreg
>            comptue m(cut+1) =
>sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>        end dofor
> else if robust == 1
> dofor m=base+1+numreg to base+2*numreg
>            comptue m(cut+1) = sqrt(%xx(m-base-numreg,m-base-numreg))
>        end dofor
>
>
>
>    end do cut
>    compute fbeta = fobs; compute lbeta = drop
>    }
>else if MOVE
>    {
>*
>* Moving sample regression
>*
>    dis 'using MOVE=' ### move
>    scratch 2*numreg fobs+move-1 lobs base
>    cmoment fobs fobs+move-1
>    # x y
>    linreg(cmom,noprint,robusterrors=robust,damp=damp,lags=lags) y
>    # x
>    dofor m=base+1 to base+numreg
>        compute m(fobs+move-1) = %beta(m-base)
>    end dofor
>    if robust == 0
>     dofor m=base+1+numreg to base+2*numreg
>        compute m(fobs+move-1) =
>sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>     end dofor
>    else if robust == 1
>  dofor m=base+1+numreg to base+2*numreg
>        compute m(fobs+move-1) = sqrt(%xx(m-base-numreg,m-base-numreg))
> end dofor
>
>    do cut=fobs+move,lobs
> cmoment(setup) cut-move+1 cut
> # x y
>        overlay xy(1,cut-fobs+1) with xtyt(numreg+1)
>        compute %cmom = %cmom + xtyt*tr(xtyt)
>        overlay xy(1,cut-move-fobs+1) with xtyt(numreg+1)
>        compute %cmom = %cmom - xtyt*tr(xtyt)
>        linreg(cmom,noprint,robusterrors=robust,damp=damp,lags=lags) y
>        # x
>        dofor m=base+1 to base+numreg
>            compute m(cut) = %beta(m-base)
>        end dofor
> if robust == 0
>         dofor m=base+1+numreg to base+2*numreg
>            compute m(cut) =
>sqrt(%xx(m-base-numreg,m-base-numreg)*%seesq)
>         end dofor
> else if robust == 1
>   dofor m=base+1+numreg to base+2*numreg
>            compute m(cut) = sqrt(%xx(m-base-numreg,m-base-numreg))
>          end dofor
>    end do cut
>    compute fbeta = fobs+move-1; compute lbeta = lobs
>    }
>*
>* Now output the results
>*
>if (add==0)+(drop==0)+(move==0) == 2
>{
>do k = 1,numreg
>    labels base+k base+k+numreg
>#  %label([series]x(k)) %concat('se',%label([series]x(k)))
>end do k
>if print
>    print(dates) fbeta lbeta base+1 to base+2*numreg
>if copy
>    copy(format=format) fbeta lbeta  base+1 to base+2*numreg
>if graph
>    {
>    scratch 2 fbeta lbeta stdbase
>    labels stdbase+1 stdbase+2
>    # '95%upper' '95%lower'
>    do k = base+1, base+numreg
>        set stdbase+1 fbeta lbeta = k + (k+numreg)*1.96
>        set stdbase+2 fbeta lbeta = k - (k+numreg)*1.96
>        if graph
>            {
>            graph(dates,header='rollreg estimated betas') 3
>            # k fbeta lbeta
>            # stdbase+1 fbeta lbeta
>            # stdbase+2 fbeta lbeta
>            }
>    end do k
>    }
>dis 'Coefficients stored in series' ### base+1 'to' ### base+numreg
>'and'
>dis 'standard errors in series' ### base+numreg+1 'to' ### base+2*numreg
>
>}
>*
>end proc rollreg
>env echo
>
> * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
>*
> *  ROLLREG.SRC
> *  Copyright 1993 by the Bank of Canada
>*
> *  Programmed for RATS by Robert Amano, Jeff Gable and Simon van Norden
>*
> *  In exchange for access to these procedures, users are requested to
>*
> *  acknowledge the use of the Bank of Canada RATS procedures in
>*
> *  all published work.
>*
> * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
>*
>
>
>





---------- End of message ----------

From: "Christopher F Baum" <baum@bc.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: ROLLREG.SRC
Date: Thu, 28 May 1998 09:46:09 -0400
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
X-Mailer: Mulberry (MacOS) [1.3.3, s/n P020-300786-009] (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

With Simon van Norden's approval, his ROLLREG.SRC rolling regressions
procedure has been posted to the Statistical Software Components archive
which I maintain on IDEAS (ideas.uqam.ca), which is part of the worldwide
RePEc project. This archive, which contains over 100 components, has
primarily Stata procedures, but welcomes source-code contributions for RATS,
GAUSS, Mathematica, and MATLAB as well as FORTRAN, C, perl, etc. There are a
few RATS procedures there currently. The IDEAS framework allows scrutiny of
the module's purpose, and a contact for the author, as well as download of
the source code from a web browser using ftp. 

The home page for the archive is

http://ideas.uqam.ca/ideas/data/bocbocode.html

Or, it may be accessed from the IDEAS homepage (http://ideas.uqam.ca) under
'software'->'Boston College'.

Two notes:

(1) the copy of ROLLREG.SRC posted to RATS-L earlier this week may contain
some inappropriate line breaks, depending on your mailer. The copy in the
archive has been tested and is in working order.

(2) You are very welcome to submit any RATS procedures you wish to share to
the SSC Archive; particularly for lengthy submissions, that is far handier
than sending them to the list. The Stata procedures on the SSC Archive have
been contributed by Stata-List participants. I imagine that there's just as
much shareable RATS code out there; the SSC Archive provides a ready
mechanism for its dissemination, and the SSC Archive is searchable by an
eXcite engine for keywords in the 'abstracts' describing each module. Test
programs and small data files used to illustrate the use of a program can
also be shared in this manner. Items are posted in latest-first order; you
may of course do a crude search in your browser with the 'find' button (e.g.
find 'rats'). If you'd like to share any of your well-documented RATS
procedures, please just email them to me at baum@bc.edu. I will be glad to
post updated/corrected versions when provided.

Thanks
Kit Baum
SSC maintainer

---------- End of message ----------

From: YPORTILL@macrocon.com.pe (Yolanda Portilla)
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: a question about maxima verosimilitud
Date: Thu, 28 May 1998 03:38:55 +0000
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
X-mailer: Pegasus Mail for Windows (v2.52) (via Mercury MTS (Bindery) v1.30)
Organization: MACROCONSULT

thank you very much for your advice.

saludos



Thank you
Yolanda Portilla Sotomayor
Economist
Pontificia Universidad Catolica del Peru

---------- End of message ----------

From: YPORTILL@macrocon.com.pe (Yolanda Portilla)
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: a question about maxima verosimilitud
Date: Thu, 28 May 1998 03:40:05 +0000
Errors-to: <rats-l-owner@efs.mq.edu.au>
Reply-to: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Sender: Maiser@efs1.efs.mq.edu.au
X-listname: <RATS-L@efs1.efs.mq.edu.au>
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
X-mailer: Pegasus Mail for Windows (v2.52) (via Mercury MTS (Bindery) v1.30)
Organization: MACROCONSULT

Thanks for your advice. 

Thank you
Yolanda Portilla Sotomayor
Economist
Pontificia Universidad Catolica del Peru

---------- End of message ----------