Return-Path: <maiser@efs.mq.edu.au>
Received: from corot.bc.edu (corot.bc.edu [136.167.2.209])
	by monet.bc.edu (8.8.7/8.8.7) with ESMTP id OAA161098
	for <BAUM@monet.bc.edu>; Fri, 1 May 1998 14:46:47 -0400
From: maiser@efs.mq.edu.au
Received: (from root@localhost)
	by corot.bc.edu (8.8.7/8.8.7) with X.500 id OAA177290
	for BAUM@MAIL1.BC.EDU; Fri, 1 May 1998 14:46:47 -0400
Received: from baldrick.ocs.mq.edu.au (baldrick.ocs.mq.edu.au [137.111.1.12])
	by corot.bc.edu (8.8.7/8.8.7) with ESMTP id OAA50516
	for <baum@bc.edu>; Fri, 1 May 1998 14:46:42 -0400
Received: from efs1.efs.mq.edu.au (efs1.efs.mq.edu.au [137.111.64.8])
	by baldrick.ocs.mq.edu.au (8.8.5/8.8.5) with ESMTP id EAA23408
	for <baum@bc.edu>; Sat, 2 May 1998 04:46:37 +1000 (EST)
Received: from EFS1/SpoolDir by efs1.efs.mq.edu.au (Mercury 1.21);
    2 May 98 04:45:48 GMT+1000
Received: from SpoolDir by EFS1 (Mercury 1.30); 2 May 98 04:45:46 GMT+1000
To: baum@bc.edu
Date: Sat, 2 May 98 4:45:46 GMT+1000
Subject: Re: 
Message-ID: <194FCDE1978@efs1.efs.mq.edu.au>

From: Luca Cazzulani <luca.cazzulani@uni-bocconi.it>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Neural networks
Date: Wed, 01 Apr 1998 18:03:56 +0100
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: Universita' Bocconi - Milano (Italy)
X-Mailer: QUALCOMM Windows Eudora Light Version 3.0.5 (32) (via Mercury MTS (Bindery) v1.30)
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"

Hi everybody!

Is anyone aware of a gauss routine to simulate a Neural Network?

Many thanks in advance
Luca Cazzulani 


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

From: P.J.Dawson@Newcastle.ac.uk (Phil Dawson)
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Zivot+Andrews Unit Root Test
Date: Thu, 2 Apr 1998 01:00:23 +0100
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

I am looking for a procedure that operationalises the Zivot & Andrews test
from Jol Bus & Econ Stats 1992 - Ho: unit root with a structural break; H1:
trend stationary with a time break. Thanks.

Dr. P.J. Dawson,
Department of Agricultural Economics and Food Marketing,
University of Newcastle upon Tyne,
Newcastle upon Tyne,
UK.
NE1 7RU.

Tel:  +44 (0)191 222 6883
Fax:  +44 (0)191 222 6720



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

From: Manfred.Kremer@bundesbank.de
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Monthly statistics on daily data
Date: 06 Apr 1998 10:30:14 +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>
X-Mailer: Mercury MTS (Bindery) v1.30


Dear RATS freaks:
Does anybody know how to use RATS for calculating time series of exact
monthly statistics on daily data, e.g., for calculating the standard
deviation of daily interest rate changes within each month, without taking
recourse to approximate solutions (say, by using the MVSTATS instruction
which calculates moving statistics over fixed intervals of time)?
Any suggestion will be highly welcome!
Ciao
Manfred



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

From: "Christopher F Baum" <baum@bc.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: Monthly statistics on daily data
Date: Mon, 06 Apr 1998 08:54:55 -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: 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

RATS is surely not the most adept program to do this, but it can be readily
done...

in a loop over the months in your dataset (this is easiest if you create an
integer variable using date functions which identifies the possible months),


create a dummy variable that is 1 when you are in the appropriate month,
then

use stats(smpl=dummy,noprint) to calculate the desired summary statistic
(e.g. %variance), and 

store it as an element of a RECT array indexed by the loop counter. 

You could calculate more than one summary statistic within the loop with
this method.

Kit Baum
Boston College

--On Mon, Apr 6, 1998 10:30 +0000 Manfred.Kremer@bundesbank.de wrote: 

> 
> Dear RATS freaks:
> Does anybody know how to use RATS for calculating time series of exact
> monthly statistics on daily data, e.g., for calculating the standard
> deviation of daily interest rate changes within each month, without taking
> recourse to approximate solutions (say, by using the MVSTATS instruction
> which calculates moving statistics over fixed intervals of time)?
> Any suggestion will be highly welcome!
> Ciao
> Manfred
> 
> 



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

From: pierpaolo grippa <me1118@mclink.it> (by way of Rob Trevor)
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Marquardt method
Date: Thu, 9 Apr 1998 09:09:10 +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

Dear rats-people

does anyone know how to implement the (Levenberg-)Marquardt method in Rats ?

I would also welcome any suggestion about alternative techniques to estimate
a CIR term structure of interest rates with daily price data on coupon bonds.

Bye.

________________
Pierpaolo Grippa
Banking Supervision Dept.
Bank of Italy
e-maill: p.grippa@mclink.it



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

From: Cem ERTUR <certur@u-bourgogne.fr>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Kalman filter
Date: Thu, 09 Apr 1998 18:57:01 +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: Windows Eudora Pro Version 3.0.1 (32) [F] (via Mercury MTS (Bindery) v1.30)
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"


Colleagues,

I would like to know how to program in Ra	ts the estimation of a model of
unobserved components : trend-cycle, seasonality, trading-day variations
and irregulars (structural time series models with explanatory variables)
presented in state space form and using Kalman filter (Harvey, 1989 and
more precisely Dagum and Quenneville in Journal of Econometrics 1993).
________________________________________
         Cem ERTUR
LATEC, Pole d'Economie et de Gestion
       2, bd Gabriel
     21000 Dijon-France
tel : (+33)-03-80-39-35-23
fax : (+33)-03-80-39-54-43
e-mail : cem.ertur@u-bourgogne.fr
_________________________________________

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

From: Simon van Norden <svn@alum.mit.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Citibase dataset conversion
Date: Thu, 16 Apr 1998 11:12:19 -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 have a few big (100+ series) data files that I want to convert to
RATS-format data files.
However, they are in FAME format (which I understand is the same as
CITIBASE.)

RATSDATA won't let me import them into a RATS database.
Is there any alternative way to convert them without reading and saving
each series individually?

--
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: Citibase dataset conversion
Date: Thu, 16 Apr 1998 10:29:43 -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 have a few big (100+ series) data files that I want to convert to
> RATS-format data files.
> However, they are in FAME format (which I understand is the same as
> CITIBASE.)
> 

FAME format is not the same as CITIBASE format. FAME is a database
management software package, which includes its own native file
format. For several years, FAME software was indeed in charge of
distributing the Citibase database, but this data was distributed in
an entirely different format than that used by their FAME database
management software.

The only way to directly access FAME database files with RATS is to 
use the UNIX version of RATS with the RATS-FAME interface add-on, 
which allows RATS to read/write FAME data files. If you don't have 
access to this environment, you will probably need either:

a) Access to the FAME software, which should allow you to write the 
data out in a format you can read in RATS (DBF, text, spreadsheet, 
etc.), or

b) Some other 3rd party software that can transfer data from FAME 
format to other formats (I don't know if any such products exist).

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: "Christopher F Baum" <baum@bc.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: Citibase dataset conversion
Date: Thu, 16 Apr 1998 11:38:17 -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 glance at the webpage of StatTransfer (www.stattransfer.com), generally
the Swiss Army knife of database conversions, indicates that they do _not_
support FAME.

Kit Baum
Boston College

--On Thu, Apr 16, 1998 10:29 -0600 Estima <estima@estima.com> wrote: 

> 
>> I have a few big (100+ series) data files that I want to convert to
>> RATS-format data files.
>> However, they are in FAME format (which I understand is the same as
>> CITIBASE.)
>> 
> 
> FAME format is not the same as CITIBASE format. FAME is a database
> management software package, which includes its own native file
> format. For several years, FAME software was indeed in charge of
> distributing the Citibase database, but this data was distributed in
> an entirely different format than that used by their FAME database
> management software.
> 
> The only way to directly access FAME database files with RATS is to 
> use the UNIX version of RATS with the RATS-FAME interface add-on, 
> which allows RATS to read/write FAME data files. If you don't have 
> access to this environment, you will probably need either:
> 
> a) Access to the FAME software, which should allow you to write the 
> data out in a format you can read in RATS (DBF, text, spreadsheet, 
> etc.), or
> 
> b) Some other 3rd party software that can transfer data from FAME 
> format to other formats (I don't know if any such products exist).
> 
> 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: richard priestley <richard.priestley@bi.no>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: 
Date: Fri, 17 Apr 1998 11:09:06 +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: Windows Eudora Light Version 3.0.1 (16) (via Mercury MTS (Bindery) v1.30)
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"

I am estimating a system using the MAXIMIZE command. I want to restrict one
of the estimated coefficient to be 0 <= b <= 1, i.e. b lies between 1 and 0.

Any ideas on how I might write this restriction in my code?

Many thanks in advance

Richard Priestley


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

From: "Frieder Knüpling" <knueplin@ruf.uni-freiburg.de>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: 
Date: Fri, 17 Apr 1998 11:37:45 +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 Richard,

you could replace b in the instruction by F(a), where F is a strictly
monotone increasing function with values between 0 and 1 (e.g. a
cumulative density, or the logistic), maximize with respect to a
(unrestricted) and compute b = F(a) after MAXIMIZE has finished.

Yours, 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: James Peery Cover <jcover@ua1vm.ua.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: Richard Priestley
Date: Fri, 17 Apr 1998 09:17:31 -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>
Organization: University of Alabama
X-Mailer: Mozilla 4.04 [en] (Win95; I) (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="------------137335061E780B81D66BE59F"

This is a multi-part message in MIME format.
--------------137335061E780B81D66BE59F
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Try b = (a/(1+a))

richard priestley wrote:

> I am estimating a system using the MAXIMIZE command. I want to restrict one
> of the estimated coefficient to be 0 <= b <= 1, i.e. b lies between 1 and 0.
>
> Any ideas on how I might write this restriction in my code?
>
> Many thanks in advance
>
> Richard Priestley



--
James Peery Cover
Department of Economics and Finance
Culverhouse College of Commerce
University of Alabama
P. O. Box 870224
Tuscaloosa, AL 35487-0224


--------------137335061E780B81D66BE59F
Content-Type: text/x-vcard; charset=us-ascii; name="vcard.vcf"
Content-Transfer-Encoding: 7bit
Content-Description: Card for James Cover
Content-Disposition: attachment; filename="vcard.vcf"

begin:          vcard
fn:             James Cover
n:              Cover;James
org:            University of Alabama
email;internet: jcover@ua1vm.ua.edu
title:          Associate Professor of Economics
x-mozilla-cpt:  ;0
x-mozilla-html: FALSE
version:        2.1
end:            vcard


--------------137335061E780B81D66BE59F--


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

From: "Frieder Knüpling" <knueplin@ruf.uni-freiburg.de>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: Re: Richard Priestley
Date: Fri, 17 Apr 1998 17:12:34 +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=iso-8859-1
Content-Transfer-Encoding: 8bit

This will not work. a/(1+a) attains negative values for -1<a<0 and values greater
than 1 for a<-1.

Frieder Knüpling


James Peery Cover wrote:

> Try b = (a/(1+a))
>
> richard priestley wrote:
>
> > I am estimating a system using the MAXIMIZE command. I want to restrict one
> > of the estimated coefficient to be 0 <= b <= 1, i.e. b lies between 1 and 0.
> >
> > Any ideas on how I might write this restriction in my code?
> >
> > Many thanks in advance
> >
> > Richard Priestley
>
> --
> James Peery Cover
> Department of Economics and Finance
> Culverhouse College of Commerce
> University of Alabama
> P. O. Box 870224
> Tuscaloosa, AL 35487-0224
>
>   ------------------------------------------------------------------------
>
>   James Cover <jcover@ua1vm.ua.edu>
>   Associate Professor of Economics
>   University of Alabama
>
>   James Cover
>   Associate Professor of Economics  <jcover@ua1vm.ua.edu>
>   University of Alabama
>                                     Netscape Conference Address
>                                     Netscape Conference DLS Server
>   Additional Information:
>   Last Name      Cover
>   First Name     James
>   Version        2.1



--
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: Benoit Carmichael <benoit@ecn.ulaval.ca>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: RE: 
Date: Fri, 17 Apr 1998 11:25:12 -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: Messagerie Internet de Microsoft/MAPI - 8.0.0.4211 (via Mercury MTS (Bindery) v1.30)

Try b=(1/(1+a**2))

Benoit Carmichael


Benoit Carmichael
Departement d'economique
Pavillon de Seve 
Universite Laval
Ste-Foy (Quebec)
G1K 7P4
Tel. : (418) 656-2131-5442
Fax : (418) 656-2707
courriel : benoit.carmichael@ecn.ulaval.ca



-----Message d'origine-----
De:	richard priestley [SMTP:richard.priestley@bi.no]
Date:	17 avril, 1998 05:09
A:	RATS Discussion List
Objet:	

I am estimating a system using the MAXIMIZE command. I want to restrict one
of the estimated coefficient to be 0 <= b <= 1, i.e. b lies between 1 and 0.

Any ideas on how I might write this restriction in my code?

Many thanks in advance

Richard Priestley


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

From: arturo jose galindo <agalindo@students.uiuc.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Newey-West
Date: Fri, 17 Apr 1998 17:39:16 -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 4.01 [en] (Win95; I) (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Hello everyone. Please correct me if I am wrong:

If I execute the following:

Linreg(Robusterrors,lags=4,damp=1) x ; # constant y  z

does the significance level reported in the regression output,
corresponds to the signif. level of a chi squared distributed assympotic
test, using a Newey-West type consistent covariance matrix? Or should I
do something different in order to run tests using such a matrix.

In advance thanks,

Arturo Galindo
Department of Economics
University of Illinois at Urbana-Champaign



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

From: jtebeka@oddo.fr
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: estimate var-cov matrix of resiuals at each step of maximize
Date: Mon, 20 Apr 1998 10:23:37 +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>
Mime-Version: 1.0
Content-type: text/plain; charset=us-ascii
X-Mailer: Mercury MTS (Bindery) v1.30




I am trying to estimate the var-cov matrix of residuals at each step of
maximize instruction. This is what i done. Althought, i think it
should work, it does not !
The message i have is the following :

Non-Linear Optimization, Function Evaluation 2.
 Old Function = 7.476038e+002   New Function = 6.234864e+003
 New Coefficients:
      0.000000       0.050000       0.050000       0.050000       0.045000


Estimation by Simplex
Monthly Data From 76:01 To 98:04
Usable Observations    268      Degrees of Freedom   263
Function Value                        NA

   Variable                     Coeff
*****************************************
1.  M                          0.0000000000
2.  VA11                     0.0500000000
3.  VA22                     0.0500000000
4.  VB11                     0.0500000000
5.  VB22                     0.0450001526

## SR10. Missing Values And/Or SMPL Options Leave No Usable Data Points

Could you please help me.
Jacques.
PS: when i fixe H0 to a constant (so i delete the lines below with ";>>>"
marks, the program works fine).

FRML LOGL = $
     H11(T)= H11F(T)   , H22(T)=H22F(T) , H12(T)=H12F(T) , $
     U1(T) = RESID1(T) , U2(T) = RESID2(T) , $
     H = ||H11(T)|H12(T),H22(T)|| , $
     U = ||U1(T),U2(T)||, $
     %LOGDENSITY(H,U)
*
NONLIN(ADD) VA11 VA22 VB11 VB22
*
comp Id = %identity(2)
FRML H11F = $
  (VA=||VA11,0.|0.,VA22||),$
  (VB=||VB11,0.|0.,VB22||),$
  (H0 = ||u1(t)|u2(t)||*||u1(t),u2(t)||),$ ;>>> var-cov calculation
  (VC=H0*(Id - (%xdiag(VA)*tr(%xdiag(VA))) - (%xdiag(VB)*tr(%xdiag(VB))))),
 $
  (H=||H11{1}|H12{1},H22{1}||),$
  (UB=||U1{1},U2{1}||*VB),$
  (H=TR(VC)*VC+%MQFORM(H,VA)+TR(UB)*UB),$
  H(1,1)
FRML H12F = H(1,2)
FRML H22F = H(2,2)
COMPUTE VA11 = VA22 = 0.05
COMPUTE VB11 = VB22 = 0.05
COMPUTE M = 0.
*
nlpar(cvcrit=10e-5,subiter=500)
MAXIMIZE(trace,METHOD=SIMPLEX,RECURSIVE,ITERS=10) LOGL GSTART GEND
MAXIMIZE(trace,METHOD=BHHH,RECURSIVE,ITERS=100) LOGL GSTART GEND



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

From: "Estima" <estima@estima.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: estimate var-cov matrix of resiuals at each step of maximize
Date: Mon, 20 Apr 1998 12:00:32 -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 trying to estimate the var-cov matrix of residuals at each step of
> maximize instruction. This is what i done. Althought, i think it
> should work, it does not !
> The message i have is the following :
> 
> Non-Linear Optimization, Function Evaluation 2.
>  Old Function = 7.476038e+002   New Function = 6.234864e+003
>  New Coefficients:
>       0.000000       0.050000       0.050000       0.050000       0.045000
> 
> 
> Estimation by Simplex
> Monthly Data From 76:01 To 98:04
> Usable Observations    268      Degrees of Freedom   263
> Function Value                        NA
> 
>    Variable                     Coeff
> *****************************************
> 1.  M                          0.0000000000
> 2.  VA11                     0.0500000000
> 3.  VA22                     0.0500000000
> 4.  VB11                     0.0500000000
> 5.  VB22                     0.0450001526
> 

First, I'm a little skeptical as to whether this setup will work, 
as I haven't seen anything quite like it tried. 

In any case, it appears that M is in your list of non-lin
parameters, but does not seem to appear in any of your formulas.
That might be the problem. Otherwise, you'll need to do some
detective work.

From the output, it appears that the SIMPLEX estimation works (at 
least in the sense it doesn't generate an error), but ends up with NA 
for the function value, indicating that the LOGL formula contains 
only NA's. 

Instead of doing the MAXIMIZE instructions, you need to use SET 
instructions to store the FRML values (given the initial conditions) 
into series, and then print those series out to see if you can find 
the source of the NA's. If everything looks OK, do the SIMPLEX 
estimation and then use SET's and PRINT's again to look at all the 
FRML values (you can also print out the various series that are set 
automatically by the FRMLs, such as U1, H11, etc.).

For example, after your COMPUTEs, do:

*
comp Id = %identity(2)
FRML H11F = $
  (VA=||VA11,0.|0.,VA22||),$
  (VB=||VB11,0.|0.,VB22||),$
  (H0 = ||u1(t)|u2(t)||*||u1(t),u2(t)||),$ ;>>> var-cov calculation
  (VC=H0*(Id - (%xdiag(VA)*tr(%xdiag(VA))) -
  (%xdiag(VB)*tr(%xdiag(VB))))),
 $
  (H=||H11{1}|H12{1},H22{1}||),$
  (UB=||U1{1},U2{1}||*VB),$
  (H=TR(VC)*VC+%MQFORM(H,VA)+TR(UB)*UB),$
  H(1,1)
FRML H12F = H(1,2)
FRML H22F = H(2,2)
COMPUTE VA11 = VA22 = 0.05

set h11fser gstart gend = h11f(t)
set h12fser gstart gend = h12f(t)
etc.

set loglser gstart gend = logl(t)
print / h11fser h12fser h22fser loglser u1 u2

and so on. Look for NA's, then work backwards figuring out which part 
of the formula in question is causing the problem.

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: Manfred.Kremer@bundesbank.de
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Newey-West
Date: 20 Apr 1998 20:32:04 +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>
X-Mailer: Mercury MTS (Bindery) v1.30


Dear Arturo:

You have asked the following question:

***************************************************************************
***************************************************************************
**************************

   Hello everyone. Please correct me if I am wrong:
   If I execute the following:
   Linreg(Robusterrors,lags=4,damp=1) x ; # constant y  z
   does the significance level reported in the regression output,
   corresponds to the signif. level of a chi squared distributed assympotic
   test, using a Newey-West type consistent covariance matrix? Or should I
   do something different in order to run tests using such a matrix.
   In advance thanks,
   Arturo Galindo
***************************************************************************
***************************************************************************
**************************

Answer: Yes, with the above instruction you get a consistent estimate of
      the standard errors of the OLS regression coefficients which corrects
      for moving-average-type autocorrelation and heteroscedasticity (of
      unknown form) in the residuals. Since the OLS regression coefficients
      can in that case be regarded as asymptotically normally distributed,
      the significance levels shown in the RATS regression output are
      derived from the normal distribution.



You can prove it as follows:



       LINREG x start end residuals
      # constant y z
      MCOV(lags=4,damp=1.,noprint) start end residuals
      # constant y z
      COM vcvnw = %XX*%CMOM*%XX                  ;** see RATS-MANUAL pp.
      14-156 ff.
      *
      COM stat1 = %beta(1)/SQRT(%SCALAR(vcvnw(1,1)))
      COM stat2 = %beta(2)/SQRT(%SCALAR(vcvnw(2,2)))
      COM stat3 = %beta(3)/SQRT(%SCALAR(vcvnw(3,3)))
      *
      CDF normal stat1
      CDF normal stat2
      CDF normal stat3
      *

       Compare the t-statistics and significance levels with those of the
      regression output when using the robusterrors option. By the way, if
      you set "damp=0", you get - if possible - the corresponding
      Hansen/Hodrick estimates (see RATS-MANUAL, pp. 14-156 ff. and, e.g.,
      HAMILTON 1994, pp.218-220).

Best regards

Manfredo




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

From: "Eric Thode" <eric.thode@mail.uni-wuerzburg.de>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: ADF Z-Test
Date: Tue, 21 Apr 1998 15:11:59 +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: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
X-mailer: Pegasus Mail for Windows (v2.53/R1) (via Mercury MTS (Bindery) v1.30)

Dear colleagues,

the RATS procedure URADF.SRC computes a statistic called "Augmented 
Dickey Fuller Z-Test". Can anybody tell me what this Z-test is all 
about? Don't get me wrong, I don't need a general introduction in 
Dickey-Fuller  unit root tests, I would just like to know what the 
properties of this Z-test are. Any hints or literature sources are 
welcome.
Thanks,

Eric
Dipl.-Vw. Eric Thode
Dept. of Economics (VWL 4)
University of Wuerzburg
Sanderring 2
97070 Wuerzburg

Tel.: +49-931-312928
Fax.: +49-931-312774

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

From: Aricio Xavier de Oliveira <Aricio@eaesp.fgvsp.br>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: RE: ADF Z-Test
Date: Tue, 21 Apr 1998 12:06:49 -0300
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 (5.0.1458.49) (via Mercury MTS (Bindery) v1.30)
Content-Type: text/plain;
Content-Transfer-Encoding: quoted-printable

Dear Eric:

References: (1) Testing for a Unit Root in Time Series Regression,
		Phillips,P.C.B. and Perron,P.
		Biometrika, 75, 335-346, 1988

	       (2) Unit Roots and Cointegration for the Economist,
		Holden,D. and Perman,R.
		in Cointegration for the Applied Economist,
		    Edited by B.Bhaskara Rao
		    Macmillan, 1994, pages 66/67

	Aricio=20




> ----------
> From: 	Eric Thode[SMTP:eric.thode@mail.uni-wuerzburg.de]
> Sent: 	Ter=E7a-feira, 21 de Abril de 1998 10:11
> To: 	RATS Discussion List
> Subject: 	ADF Z-Test
>=20
> Dear colleagues,
>=20
> the RATS procedure URADF.SRC computes a statistic called "Augmented=20
> Dickey Fuller Z-Test". Can anybody tell me what this Z-test is all=20
> about? Don't get me wrong, I don't need a general introduction in=20
> Dickey-Fuller  unit root tests, I would just like to know what the=20
> properties of this Z-test are. Any hints or literature sources are=20
> welcome.
> Thanks,
>=20
> Eric
> Dipl.-Vw. Eric Thode
> Dept. of Economics (VWL 4)
> University of Wuerzburg
> Sanderring 2
> 97070 Wuerzburg
>=20
> Tel.: +49-931-312928
> Fax.: +49-931-312774
>=20

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

From: "Coleman, Mark" <MColeman@dlj.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Survival/Hazard model in RATS
Date: Thu, 23 Apr 1998 17:05:01 -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,

Does anyone happen to have a RATS proc that estimates a proportional
hazard model or some similar survival model.

Thanks

Mark




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

From: Junsoo Lee <Leej@Ctrvax.Vanderbilt.Edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: unsubscribe
Date: Thu, 23 Apr 1998 16:32:11 -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: Windows Eudora Pro Version 3.0 (32) (via Mercury MTS (Bindery) v1.30)
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"

unsubscribe RATS
     Junsoo Lee
     P.O. Box 1698 B              615-322-3428 (voice & fax)
     Department of Economics      615-343-8495 (dept. fax)
     Vanderbilt University        615-646-9852 (home)
     Nashville, TN 37235, USA     Junsoo.Lee@Vanderbilt.Edu (e-mail) 

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

From: "903953 GIANI MARCO"<903953_GIANI_MARCO.DG@mailbox.enel.it>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: unsubscribe
Date: Fri, 24 Apr 1998 08:39:13 +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: text/plain; charset=us-ascii
X-Mailer: Mercury MTS (Bindery) v1.30





unsubscribe RATS



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

From: Ying-Man Chung <y_mchung@yahoo.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: unsubscribe 
Date: Sun, 26 Apr 1998 11:34:44 -0700 (PDT)
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 RATS



_________________________________________________________
DO YOU YAHOO!?
Get your free @yahoo.com address at http://mail.yahoo.com


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

From: Tapio Pekkala <k24255@kyyppari.hkkk.fi>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: unsubscribe
Date: Mon, 27 Apr 1998 13:03:54 +0300 (EET DST)
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 RATS



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

From: Tapio Pekkala <k24255@kyyppari.hkkk.fi>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: unsubscribe
Date: Mon, 27 Apr 1998 14:06:55 +0300 (EET DST)
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 RATS

  
 


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

From: "Estima" <estima@estima.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: unsubscribe -- Tapio Pekkala
Date: Mon, 27 Apr 1998 09:28:50 -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)


> 
> 
>  unsubscribe RATS

Quit spamming the list with your unsubscribe requests--read the 
message you were sent when you signed up, or the message on our web 
page, for information on unsubscribing.

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: Graham Smith <GS8@soas.ac.uk>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Heteroscedasticity Robust Variance Ratios
Date: Wed, 29 Apr 1998 12:13:01 GMT
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:  S.O.A.S.
X-mailer: Pegasus Mail v3.22 (via Mercury MTS (Bindery) v1.30)

Does anyone have a procedure for the heteroscedasticity-robust 
variance ratio statistics of Lo and MacKinlay, Review of Financial 
Studies 1988?

If so, please send me a copy. Thanks.

Graham Smith

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

From: "Wai Lee"<lee_wai@jpmorgan.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: panel vs SUR
Date: Wed, 29 Apr 1998 08:30:30 -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


What is the difference of using SUR with EQUATE and using cross
section/panel estimation of the same set of equations?  Both force the
coefficients to be the same across equations, and both are estimated by
GLS.  Are they really the same?

Thanks for any thoughts.

W. Lee



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

From: "=?ISO-8859-1?Q?Emilio_Caminero_Mart=EDn?=" <ecaminero@ikt.es>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: 
Date: Wed, 29 Apr 1998 15:06:36 +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: Microsoft Internet Mail 4.70.1155 (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit



unsubscribe

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

From: "=?ISO-8859-1?Q?Emilio_Caminero_Mart=EDn?=" <ecaminero@ikt.es>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: 
Date: Wed, 29 Apr 1998 15:07:16 +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: Microsoft Internet Mail 4.70.1155 (via Mercury MTS (Bindery) v1.30)
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit



unsubscribe  ecm@blpa00.bl.ehu.es

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

From: "Christopher F Baum" <baum@bc.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: panel vs SUR
Date: Wed, 29 Apr 1998 09:48:47 -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

If by 'c-s/panel estimation' you mean the fixed effects model with dummies
allowing for all slopes to vary across cross-sectional units, you are not
describing a GLS technique. An obvious difference between pooled estimation
methods and SUR (with or without equate) is that the former estimate a
single error variance for the entire panel. Unless you have verified that
the errors are homoscedastic over i, that would be an inappropriate
constraint. SUR also allows for the contemporaneous correlation of e(i,t)
and e(j,t), which is assumed to be zero in an OLS application. If it is not
(i.e. if SUR does any good) then OLS is not appropriate, as OLS estimation
assumes zero covariance between every pair of errors.

Kit Baum
Boston College

--On Wed, Apr 29, 1998 8:30 -0400 "Wai Lee" <lee_wai@jpmorgan.com> wrote: 

> 
> What is the difference of using SUR with EQUATE and using cross
> section/panel estimation of the same set of equations?  Both force the
> coefficients to be the same across equations, and both are estimated by
> GLS.  Are they really the same?
> 
> Thanks for any thoughts.
> 
> W. Lee
> 
> 



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

From: "Wai Lee"<lee_wai@jpmorgan.com>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: panel vs SUR
Date: Wed, 29 Apr 1998 10:15:16 -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


Thanks for sharing your thoughts.

I understand the advantage of SUR over OLS when it is used appropriately.
I guess what I am not sure is on the capability of the c-s/panel estimation
with "random effects" in RATS.  I thought we could impose cross-sectional
contemporaneous correlation, as well as different variances across
equations.  May be I should check the Hausman and Taylor (1981) as cited in
RATS manual.

W. Lee






baum @ bc.edu on 04/29/98 09:48:47 AM

Please respond to RATS-L@efs.mq.edu.au

To:   RATS-L @ efs.mq.edu.au
cc:   (Wai Lee)
Subject:  Re: panel vs SUR





If by 'c-s/panel estimation' you mean the fixed effects model with dummies
allowing for all slopes to vary across cross-sectional units, you are not
describing a GLS technique. An obvious difference between pooled estimation
methods and SUR (with or without equate) is that the former estimate a
single error variance for the entire panel. Unless you have verified that
the errors are homoscedastic over i, that would be an inappropriate
constraint. SUR also allows for the contemporaneous correlation of e(i,t)
and e(j,t), which is assumed to be zero in an OLS application. If it is not
(i.e. if SUR does any good) then OLS is not appropriate, as OLS estimation
assumes zero covariance between every pair of errors.
Kit Baum
Boston College
--On Wed, Apr 29, 1998 8:30 -0400 "Wai Lee" <lee_wai@jpmorgan.com> wrote:
>
> What is the difference of using SUR with EQUATE and using cross
> section/panel estimation of the same set of equations?  Both force the
> coefficients to be the same across equations, and both are estimated by
> GLS.  Are they really the same?
>
> Thanks for any thoughts.
>
> W. Lee
>
>






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

From: "Christopher F Baum" <baum@bc.edu>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: panel vs SUR
Date: Wed, 29 Apr 1998 10:25:01 -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

One must be very careful with the random effects estimator, for which E(X'e)
must be zero for consistency (that is the purpose of the Hausman test to
compare fixed vs. random effects). RATS is much less developed at doing this
sort of estimation than is, for example, Stata with its 'xt' command set.

Kit Baum

--On Wed, Apr 29, 1998 10:15 -0400 "Wai Lee" <lee_wai@jpmorgan.com> wrote: 

> 
> Thanks for sharing your thoughts.
> 
> I understand the advantage of SUR over OLS when it is used appropriately.
> I guess what I am not sure is on the capability of the c-s/panel
estimation
> with "random effects" in RATS.  I thought we could impose cross-sectional
> contemporaneous correlation, as well as different variances across
> equations.  May be I should check the Hausman and Taylor (1981) as cited
in
> RATS manual.
> 
> W. Lee
> 
> 
> 
> 
> 
> 
> baum @ bc.edu on 04/29/98 09:48:47 AM
> 
> Please respond to RATS-L@efs.mq.edu.au
> 
> To:   RATS-L @ efs.mq.edu.au
> cc:   (Wai Lee)
> Subject:  Re: panel vs SUR
> 
> 
> 
> 
> 
> If by 'c-s/panel estimation' you mean the fixed effects model with dummies
> allowing for all slopes to vary across cross-sectional units, you are not
> describing a GLS technique. An obvious difference between pooled
estimation
> methods and SUR (with or without equate) is that the former estimate a
> single error variance for the entire panel. Unless you have verified that
> the errors are homoscedastic over i, that would be an inappropriate
> constraint. SUR also allows for the contemporaneous correlation of e(i,t)
> and e(j,t), which is assumed to be zero in an OLS application. If it is
not
> (i.e. if SUR does any good) then OLS is not appropriate, as OLS estimation
> assumes zero covariance between every pair of errors.
> Kit Baum
> Boston College
> --On Wed, Apr 29, 1998 8:30 -0400 "Wai Lee" <lee_wai@jpmorgan.com> wrote:
>>
>> What is the difference of using SUR with EQUATE and using cross
>> section/panel estimation of the same set of equations?  Both force the
>> coefficients to be the same across equations, and both are estimated by
>> GLS.  Are they really the same?
>>
>> Thanks for any thoughts.
>>
>> W. Lee
>>
>>
> 
> 
> 
> 
> 



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

From: "Klaus P. Fischer" <Klaus.Fischer@fas.ulaval.ca>
To: "RATS Discussion List" <RATS-L@efs.mq.edu.au>
Subject: Re: Survival/Hazard model in RATS
Date: Sat, 25 Apr 1998 02:56:27 -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>
Organization: FOGAFIN/Universite Laval
X-Mailer: Mozilla 4.01 [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 Mark:

Below is a program (only the equations definitions and estimation
command) that estimates a  Split-Population Hazard Model according to
Rebel A.Cole, Jeffery W.Gunther, 'SEPARATING THE LIKELIHOOD AND TIMING
OF  BANK FAILURE', Journal of Banking & Finance 19, 1995, pp. 1073-1089.
This can give you a guide how to write simpler (or more complex) models.
You may also look at "Applied Logistic Regressions" by Hosmer and
Lemeshow (Wiley & Sons), 1989, for estimation of hazard models using
logistic regression (Logit).

set D = 0.01
set L = 0.01
set Ti = 0.01
set S = 0.01

Nonlin A1 A2 A3 A4 A5 B1 B2 B3 B4 B5
Frml Delta = 1/(1+(Exp(A1*var+A2*var+A3*var+A4*var)))
Frml Lambda = Exp(-1(B1*var+B2*var+B3*var+B4*var))
Frml TiFail = (L(T)*P*((L(T)*Time)**(P-1)))/((1+((L(T)*Time)**P)**2)
Frml Survival = (1/(1+((L(T)*Time)**P)))
Frml Likely = (L(T)=Lambda(T)), (D(T)=Delta(T)), (S(T)=Survival(T)),
(Ti(T)=TiFail(T)),
(((D(T)*Ti(T))**Status)*(((1-D(T))+(D(T)*S(T)))**(1-Status)))
Compute A1=0.0, A2=0.0, A3=0.0, A4=0.0, B1=0.0, B2=0.0, B3=0.0, B4=0.0,
P=0.0
NLPAR(SUBITERATIONS=5000)
MAXIMIZE(METHOD=BHHH,RECURSIVE,ITERS=5000) L 1 87

Coleman, Mark wrote:

> Greetings,
>
> Does anyone happen to have a RATS proc that estimates a proportional
> hazard model or some similar survival model.
>
> Thanks
>
> Mark



--
Klaus P. Fischer, Ph.D.
Centre de recherche en économie et finance appliquée (CRÉFA)
Faculté des sciences de l´administration
Université Laval, Québec G1K 7P4, CANADA

On sabbatical at
FOGAFIN
Carrera 7, No 35-40
Santafé de Bogotá, COLOMBIA
Ph.  51-1-285-7373
Fax 51-1-285-8587



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