From: Rob Trevor To: "RATS Discussion List" Subject: Loop Timing ? Date: Fri, 1 Mar 1996 17:28:06 +1100 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Mercury MTS v1.21 Hi Anybody tried to do any loop timing in developing RATS code? I've just been trying to work out how to find the expensive parts of some very intensive monte carlo code, and I can't quite see how to do timing tests without cutting the job up into very small bits. That is, there doesn't seem to be a function which returns the system time, and since rats buffers its screen output trying to find the elapsed time between different DIS statements won't work either. (I have considered trying to manipulate the output of SEED, but I was wondering if there's an easier way.) All ideas welcome. Thanks Rob ----------------------------------------------------------------------- Rob Trevor Associate Professor Centre for Studies in Money, Banking and Finance Macquarie University, NSW, AUSTRALIA 2109 Internet: robt@efs.mq.edu.au VOICE: +61 (2) 850-8539 FAX: +61 (2) 850-7281 or +61 (2) 418-1847 ----------------------------------------------------------------------- ---------- End of message ---------- From: Sune Karlsson To: "RATS Discussion List" Subject: Re: Loop Timing ? Date: Fri, 01 Mar 1996 09:46:48 +0100 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Organization: Stockholm School of Economics MIME-version: 1.0 X-Mailer: Mozilla 2.0 (WinNT; I) (via Mercury MTS v1.21) Content-type: text/plain; charset=us-ascii Content-transfer-encoding: 7bit Rob Trevor wrote: > > Hi > > Anybody tried to do any loop timing in developing RATS code? > > I've just been trying to work out how to find the expensive parts of some > very intensive monte carlo code, and I can't quite see how to do timing > tests without cutting the job up into very small bits. > > That is, there doesn't seem to be a function which returns the system time, > and since rats buffers its screen output trying to find the elapsed time > between different DIS statements won't work either. (I have considered > trying to manipulate the output of SEED, but I was wondering if there's an > easier way.) > > All ideas welcome. > I too have searched in wain for an instruction returning the system time. I'd hate to time the execution of a program by watching the screen with stop watch in hand, but I think that it is doable. You can turn off buffering of the screen output using the File, Preferences menu choice. -- Sune ---------------------------------------------------------------------- Sune Karlsson Internet: STSK@HHS.SE Stockholm School of Economics Phone: + 46 8 736 92 39 Box 6501, 113 83 Stockholm, Sweden Fax: + 46 8 34 81 61 ---------- End of message ---------- From: gagnonl@qsilver.queensu.ca (Louis Gagnon) To: "RATS Discussion List" Subject: Time in RATS Date: Fri, 1 Mar 1996 19:59:31 -0500 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: (via Mercury MTS v1.21) Dear Rob: I too have been stumbling into this problem onn many occasions and, finally, I succumbed to the temptation to ask Tom Maycock who wrote back to me the following: " RATS currently doesn't have a function to return the system time. We've had several requests for that, and may be able to add this feature in a future release. In the meantime, I can only think of a couple of possible workarounds. One would be to run the DOS version in batch mode, and using the DOS TIME function to return the time before and after the RATS session. Of course, would only let you determine the time required to run the entire program, not just the estimation itself. Another approach would be to do something like: OPEN COPY FIRST.TXT DISPLAY 'Starting Estimation' CLOSE COPY É. do estimation here OPEN COPY SECOND.TXT DISPLAY 'Finished with Estimation' CLOSE COPY Once the program is done, you could look at the time stamps on FIRST.TXT and SECOND.TXT to determine the approximate time required. If you are doing several estimations in one program, you could use a different pair of filenames for each estimation. Hope this is helpful, and good luck. Sincerely, Tom Maycock Estima" I hope that this helps you. At any rate, I am happy to share this information with you and, indeed, the entire discusion group. Regards, Louis Gagnon Louis Gagnon School of Business Queen's University Kingston, Ontario K7L 3N6 Business: (613) 545-6707 Facsimile: (613) 545-6589 Home: (613) 634-1837 ---------- End of message ---------- From: Rob Trevor To: "RATS Discussion List" Subject: Archives Date: Tue, 5 Mar 1996 16:38:11 +1100 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Mercury MTS v1.21 Hi all, Just a quick note to say that I believe that we've now got the archiving function tied down. While there's not much there to date, the instructions for fetching the archives are given below. (Please note that while the archive starts in Jan 96, it consists mainly of my tests of the software prior to announcing the creation of the list on Jan 25.) ------------------------------ The RATS-L list archives must be retrieved by name. RATS-L archive files are of the form RATS~y~m.ARC, where the ~y is replaced by the last two digits of the year, and the ~m is the number of the current month - ranging from 01 to 12. The list started in January 1996. To retrieve the archive for January 1996, you would send the following command to MAISER@efs.mq.edu.au in the body of the email message: SEND RATS9601.ARC ------------------------------- Rob ----------------------------------------------------------------------- Rob Trevor Associate Professor Centre for Studies in Money, Banking and Finance Macquarie University, NSW, AUSTRALIA 2109 Internet: robt@efs.mq.edu.au VOICE: +61 (2) 850-8539 FAX: +61 (2) 850-7281 or +61 (2) 418-1847 ----------------------------------------------------------------------- ---------- End of message ---------- From: nmorin@weber.ucsd.edu (Norman Morin) To: "RATS Discussion List" Subject: Nonparametric kernel estimators Date: Mon, 4 Mar 1996 22:51:33 -0800 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Mercury MTS v1.21 Hello. Does anyone know of any procedures to do nonparametric estimation of distributions? I have Gunnar Bardsen's DENSITY.SRC and I wrote KERNEL.SRC on RATS' website, but I was looking for one in which the 'window width' is also determined by the data -- cross validation techniques -- beyond just choosing the optimal Gaussian window width. Thanks alot, --Norm -- Norman J. Morin nmorin@weber.ucsd.edu ------------------------------------------------------------- Department of Economics * University of California, San Diego 9500 Gilman Drive * La Jolla, CA 92023-0508 ------------------------------------------------------------- "You have to consider we're only made out of dust. That's admittedly not much to go on and we shouldn't forget that." -- Philip K. Dick ---------- End of message ---------- From: Rob Trevor To: "RATS Discussion List" Subject: Re: Nonparametric kernel estimators Date: Tue, 5 Mar 1996 18:36:52 +1100 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Mercury MTS v1.21 At 10:51 PM 4/3/96, Norman Morin wrote: >Hello. Does anyone know of any procedures to do nonparametric >estimation of distributions? I have Gunnar Bardsen's DENSITY.SRC >and I wrote KERNEL.SRC on RATS' website, but I was looking >for one in which the 'window width' is also determined by >the data -- cross validation techniques -- beyond just choosing >the optimal Gaussian window width. > Well, the attached *may* be useful. However, PLEASE note that it is NOT good RATS code and is hence very slow. (If I wasn't trying to promote discussion on this list, I wouldn't post it since it is certainly *not* code of which I'm very proud.) I simply took some FORTRAN code and 'ratisified' it to get the job done for something I needed to do quite quickly several years ago. I had intended to go back and rewrite it properly, but É. What is needed to make this thing more useable is to take the hand-coded Newton-Raphson stuff out and use RATS's MAXIMISE instruction. (Of course, the thing really needs to be redesigned and rewritten for v 4.) Nonetheless, the code should work in the current version of RATS if you'd prefer to go and get a very long cup of coffee rather than recode the beast! Needless to say, I'd appreciate a copy of any code anyone produces based on this proc. Cheers Rob ****************************************************************************** * This procedure calculates the Gaussian kernel density estimate for the * * Sheather-Jones "solve-the-equation" bandwidth. * * * * @KERNEL( options ) SERIES NPOINTS START END * * * * Options: * * MINIMUM = Minimum value at which to estimate frequency * * MAXIMUM = Maximum value at which to estimate frequency * * SHEATHER/[NOSHEATHER] Whether want Sheather- Jones bandwidth or not * * * * Input: * * SERIES = Series containing data * * NPOINTS = Number of points at which to estimate the density * * START = First observation * * END = Last observation * * * * Output: (GLOBAL series of length NPOINTS+1) * * VALUES = Points at which density is estimated * * ESTDENS = Density estimates * * * * Note: Code based on FORTRAN source supplied by Simon Sheather * * * * Version 2.0 * * Updated for Version 4 * * July 1993 * * * * Written by Robert G Trevor * * Centre for Studies in Money, Banking and Finance * * Macquarie University, NSW, AUSTRALIA 2109 * * Internet: robt@efs.mq.edu.au * * VOICE: +61 (2) 850-8539 FAX: +61 (2) 850-7281 * * * * Reference: Sheather, S.J. and M.C. Jones. (1991). "A Reliable * * Data-based Bandwidth Selection Method for Kernel Density * * Estimation." Journal of the Royal Statistical Society, * * Series B, Vol. 53, pp 683-690 * * * ****************************************************************************** PROC KERNEL SERIES NPOINTS NSTART NEND TYPE SERIES SERIES TYPE INTEGER NSTART NEND NPOINTS OPTION REAL MINIMUM 0.0 OPTION REAL MAXIMUM 0.0 OPTION SWITCH SHEATHER 0 LOCAL SERIES DATA LOCAL INTEGER N LOOP FIRSTO LASTO START END LOCAL REAL XLO XHI LAMBDA HSJD HNML CHAT CHAT7 X DELX LOCAL REAL RT2PI L2 AC BC SA SB DA DB EA EB LOCAL REAL A ANEW C C7 TOL H Y Y2 E S TT XN LOCAL REAL K1 K2 K3 G GPRIME THREEN LOCAL REAL RHAT2 RHAT3 FIRSTH AINIT LOCAL REAL TOLA TOLG CUMMULAT TERM * EVAL XLO = MINIMUM EVAL XHI = MAXIMUM * set up convergence criteria EVAL TOLA=1.0E-5 EVAL TOLG=1.0E-4 IEVAL LOOP=15 * get sample range INQUIRE(SERIES=SERIES) START>>NSTART END>>NEND SET DATA START END = SERIES(T) * sort data low -> high ORDER DATA * get range and number of non-missing obs EXTREMUM(PRINT) DATA IF XLO.EQ.0.0 { EVAL XLO = %MINIMUM } IF XHI.EQ.0.0 { EVAL XHI = %MAXIMUM } IEVAL FIRSTO = %MINENT IEVAL LASTO = %MAXENT IEVAL N = LASTO - FIRSTO + 1 * check more obs than evaluation points IF N To: "RATS Discussion List" Subject: GMM in RATS Date: Thu, 7 Mar 1996 11:14:59 -0500 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Windows Eudora Pro Version 2.1.2 (via Mercury MTS v1.21) Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Has anyone had done GMM for a system of n equations, but each equation has different list of instruments in RATS? For example, I have a system of 7 equations. Each equation has 5 instruments. Of these 5 instruments, only the constant is common to the 7 equations. In the way generalized instrumental variable estimation is done in RATS, you can only have one INSTRUMENTS statement for the whole system (am I right?). This forces all equations to share the same list of instruments. Two years ago, Tom Maycock suggested the following method to deal with my problem. Just list all the instruments (in the e.g. above, there will be 29 of them) in the same INSTRUMENTS statement. Then use a 29x29 matrix with 0 and 1 to distinguish which instruments should enter which equation, and proceed the estimation with some modifications in the weighting matrix. I tried that but the estimates were totally off the wall, far from being reasonable. I suspect that the problem is due to the high dimension of the matrix. I finally had to switch to some matrix algebra software for the estimation. I appreciate if someone has any suggestions to do the estimation in RATS. **************************************************** Wai Lee (WLEE@HBS.EDU) Morgan Hall 362C Phone: 617-495-6368 Harvard Business School Fax: 617-496-6592 Boston, MA 02163 **************************************************** ---------- End of message ---------- From: alshebel@ksu.ksu.edu To: "RATS Discussion List" Subject: Contemporaneous test in SUR Date: Thu, 14 Mar 1996 08:47:55 +0000 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-mailer: Pegasus Mail for Windows (v2.10) (via Mercury MTS v1.21) If any one has a rats procedure (SRC) or program for computing the Lagrange multiplier statistic, suggested by Breusch and Pagan (1980)(JUDGE et al. p. 456, 1988), send me a copy. In the below example, how to compute the squared correlation between equations (residual of each) Example: EQUATION s1 q1 # CONSTANT p1 y EQUATION s2 q2 # CONSTANT p2 y EQUATION s3 q3 # CONSTANT p3 y * * LINREG(EQUATION=s1) q1 /resq1 LINREG(EQUATION=s2) q2 /resq2 LINREG(EQUATION=s3) q3 /resq3 * vcv(matrix=v) 60:1 89:1 # resq1 resq2 resq3 compute r= ? disply ' Squard Corr =' r * SUR 3 # s1 # s2 # s3 Abdullah Alshebel Kansas State University alshebel@ksu.ksu.edu ---------- End of message ---------- From: "Christopher F. Baum" To: "RATS Discussion List" Subject: Re: Contemporaneous test in SUR Date: Thu, 14 Mar 1996 13:29:48 -0400 (EDT) Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: MIME-version: 1.0 Content-type: TEXT/PLAIN; CHARSET=US-ASCII Content-transfer-encoding: 7BIT X-Mailer: Mercury MTS v1.21 Abdullah Alshebel wants to calc the LM statistic to test whether SUR is warranted (i.e. whether the covariance matrix of equation errors is diagonal). Below that derivation in Judge, the likelihood ratio test is presented, which is very straightforward--merely T ln (det(sigols) / det(sigsur)), where sigols is the var-cov mtx he calcs as v from OLS resid vectors (the constrained cov mtx) and sigsur is the var-cov mtx of the SUR (whose det is fetchable after SUR). This test is asy equiv to the LM test of same hypothesis, and under the null is distrib chi-sqr with (m)(m-1)/2 d.f. where m==number of eqns. Kit Baum Boston College ---------- End of message ---------- From: "Christopher F. Baum" To: "RATS Discussion List" Subject: Re: Contemporaneous test in SUR Date: Thu, 14 Mar 1996 13:34:52 -0400 (EDT) Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: MIME-version: 1.0 Content-type: TEXT/PLAIN; CHARSET=US-ASCII Content-transfer-encoding: 7BIT X-Mailer: Mercury MTS v1.21 Correction to my earlier posting re testing for diagonality in SUR: (1) the mtx sigols should be constrained to be diagonal, i.e. should only contain the variances of the estimated OLS residuals. (2) strictly speaking the sigsur matrix should be a maximum likelihood e estimate, which can be achieved by iterating on the cov mtx i.e. sur(iter=nn) in the SUR estimation. Kit Baum (Info from Judge Thy & Practice, 2d ed., p.476-77) ---------- End of message ---------- From: lmahens@vnet3.vub.ac.be (Luc M.A. Hens) To: "RATS Discussion List" Subject: Problems with (Classroom) RATS Date: Wed, 20 Mar 1996 16:14:13 +0200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Mercury MTS v1.21 I know that this list is focused on problems of research using RATS, but some of you may be able to help me with the following problem. I use Classroom RATS in my Econometrics course (Mac versions and Window versions). Several of my students who use Windows 95 and Windows 3.1 have been complaining that RATS continuously crashes. I have personally witnessed two cases: - one on Windows 95, where RATS crashes about every two minutes while it is in interactive mode; - one in Windows 3.1, where RATS crashes in what seem to be regular intervals of about 10 minutes (again in interactive mode). In the crash, the RATS program suddenly quits. I personally use MacRATS 4.02 and I have not had any of these problems. Has anyone out there had similar problems and knows how to fix them? ( I am aware of S. S. Gill's message of 8 February reporting similar problems under Windows 95; for Windows 95, it may be an example of its "plug and pray"-feature). Thanks in advance, Luc Hens ============================================= Luc M.A. Hens assistant professor of economics Vesalius College, Vrije Universiteit Brussel Pleinlaan 2, B-1050 Brussels Belgium phone +32-2-629 20 61 fax +32-2-629 20 60 e-mail lmahens@vnet3.vub.ac.be ============================================= ---------- End of message ---------- From: RAY DEGENNARO To: "RATS Discussion List" Subject: Re: Problems with (Classroom) RATS Date: Wed, 20 Mar 96 10:41:30 LCL Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Mercury MTS v1.21 On Wed, 20 Mar 1996 16:14:13 +0200 Luc M.A. Hens said: >Several of my students who use Windows 95 and Windows 3.1 have >been complaining that RATS continuously crashesÉ. > I have no experience with windows 95, but I do know that RATS runs perfectly under OS/2. I run multiple batch sessions without a hitch. I have estimated three arch models with 41 parameters and 35110 observations EACH running in the backgound, hooked to the mainframe for a SAS job, sent mail, and used my wordprocesser, all simultaneously with no problem. Of course, stuff that big takes a VERY long time, but I can still get my other work done in the foreground. If you want a smooth 32-bit system for RATS, I'd recommend OS/2. Ray DeGennaro ---------- End of message ---------- From: Abhay H Abhyankar To: "RATS Discussion List" Subject: RATS and Windows95/Windows NT Date: Wed, 20 Mar 1996 17:02:11 +0000 (GMT) Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Mailer: Mercury MTS v1.21 Hi: Would it be possible for someone who has used RATS with Windows95 and Windows NT to indicate the best way to load the package and also describe details of any "fiddly bits" required to get the package to run "optimally" under these two O/S's. Many Thanks A Abhyankar *~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~* | | | Dr.A Abhyankar Voice: (01786) 467287 | | Department of Accountancy and Finance Fax: (01786) 467308 | | University of Stirling | | Stirling e-mail: | | FK9 4LA aha1@stirling.ac.uk | | United Kingdom | | | *~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~*~* ---------- End of message ---------- From: Rob Trevor To: "RATS Discussion List" Subject: Compiling RATS under UNIX Date: Thu, 21 Mar 1996 10:20:05 +1100 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Mercury MTS v1.21 Hi While all these OS-type questions (and hopefully answers!) are circulating the list, I have one of my own. Does anyone (that includes Estima!) have any experience in compiling Unix RATS with optimisation turned on? The Makefile ships without any optimisation specified in the compile options. While RATS seems to compile fine under increasing levels of optimisation (at least in AIX on RS6000's), that does *not* necessarily mean that it runs faster. I haven't had time to test it, but I am faced with running a number of jobs which take large amounts of time (lots of loops involving non-linear estimations) even on decent workstations. So I'm more than ready to learn from others' experiences if it saves me doing the testing myself. Cheers Rob ----------------------------------------------------------------------- Rob Trevor Associate Professor Centre for Studies in Money, Banking and Finance Macquarie University, NSW, AUSTRALIA 2109 Internet: robt@efs.mq.edu.au VOICE: +61 (2) 850-8539 FAX: +61 (2) 850-7281 or +61 (2) 418-1847 ----------------------------------------------------------------------- ---------- End of message ---------- From: Sune Karlsson To: "RATS Discussion List" Subject: Re: RATS and Windows95/Windows NT Date: Thu, 21 Mar 1996 10:12:32 +0100 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Organization: Stockholm School of Economics MIME-version: 1.0 X-Mailer: Mozilla 2.0 (WinNT; I) (via Mercury MTS v1.21) Content-type: text/plain; charset=us-ascii Content-transfer-encoding: 7bit Abhay H Abhyankar wrote: > > Hi: > > Would it be possible for someone who has used RATS with Windows95 and > Windows NT to indicate the best way to load the package and also > describe details of any "fiddly bits" required to get the package to run > "optimally" under these two O/S's. > In general Rats 4.20 (Rats386 and WinRats-32) runs well under NT. The two exceptions I've found is that WinRats has the tedious habit of checking the A: drive (diskette) when you set the default directory (File,Directory) and Citibase directory (File,Preferences), then telling me that I don't have a diskette there. I knew that! The second (more serious) problem is that WinRats some times crash with a divide by zero error when changing directory in the default directory dialog box. Sune -- ---------------------------------------------------------------------- Sune Karlsson Internet: STSK@HHS.SE Stockholm School of Economics Phone: + 46 8 736 92 39 Box 6501, 113 83 Stockholm, Sweden Fax: + 46 8 34 81 61 ---------- End of message ---------- From: wlee@pop.hbs.edu To: "RATS Discussion List" Subject: Kalman filter Date: Thu, 21 Mar 1996 09:31:39 -0500 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Windows Eudora Pro Version 2.1.2 (via Mercury MTS v1.21) Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" In the example on p.13-6, section 13.4 of Version 4's manual, how can we estimate the autoregressive parameter of beta (pi8) and get its standard error? How can we relax the random walk version of beta and estimate pi8, instead of assigning a value of 1.0 or 0.0 to it as in the TVARYING.PRG? Thanks for any suggestions. **************************************************** Wai Lee (WLEE@HBS.EDU) Morgan Hall 362C Phone: 617-495-6368 Harvard Business School Fax: 617-496-6592 Boston, MA 02163 **************************************************** ---------- End of message ---------- From: "LUMB, IAN" To: "RATS Discussion List" Subject: Multiple Graphs with SPGRAPH & Winrats32 for NT Date: Thu, 21 Mar 96 14:41:44 GMT Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Mercury MTS v1.21 Hi, I'm having problems printing 3 graphs to a page (spgraph vfields=3) in portrait format with WinRats. Whatever I do, WinRats assumes the page format is landscape and scales the 3 charts to about 50% the size they would be if they streched from the top to the bottom of the page. My printer is setup to print Portrait. I have tried both fixed proportioning and stretch to fit on the graphics window. This wasn't a problem with Rats386 since all you needed to do was alter the orientation of the graphs (with GRPARM(PORT)), unfortunately with WinRats this corrupts the x-axes labelling. Thanks in advance for any suggestions, Ian ____________________________________________________________________ Ian Lumb ilumb@ccmail-uk.pwcm.com Research Analyst TEL: (44 171) 422 2903 Fixed Income Research FAX: (44 171) 422 2393 Painewebber International (UK) Ltd ---------- End of message ---------- From: estima@estima.com (Estima) To: "RATS Discussion List" Subject: Re: Windows 95 and Classroom WinRATS Date: Thu, 21 Mar 1996 13:12:15 -0600 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Windows Eudora Version 1.4.4 (via Mercury MTS v1.21) Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Regarding Classroom WinRATS/Windows95: 1) The Classroom version of WinRATS is labeled Version 4.21, and is thus newer than the standard WinRATS and WinRATS-32 4.20 products we are currently shipping. We will be release a Version 4.21 of these products (and the DOS/UNIX) versions shortly. The Classroom WinRATS was rushed out the door because several sites wanted an up to date Classroom version, but needed it in time for the start of 2nd semester classes. As a result, the product wasn't as fully tested as we would have liked. Comments/problems with Classroom WinRATS may therefore not apply to WinRATS 4.20. 2) There does appear to be one problem with Classroom WinRATS relating to Windows95. The program crashes fairly regularly if you do more than a couple of graphs. We hope to have this problem resolved as soon as possible. As far as we can tell, this only occurs under Windows95, and is currently the only Windows95-specific problem with RATS that we are aware of. 3) If you are experiencing crashes not related to the graphics problem noted above, please let us know, with as much detail as possible, because we can't fix 'em if we don't know about 'em! Thanks, Tom Maycock 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://homepage.interaccess.com/~estima/ | | 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: Can Ersoz To: "RATS Discussion List" Subject: Re: Kalman filter (and a question,too) Date: Thu, 21 Mar 1996 20:09:35 -0600 (CST) Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Mailer: Mercury MTS v1.21 On Thu, 21 Mar 1996 wlee@spica.hbs.edu wrote: > In the example on p.13-6, section 13.4 of Version 4's manual, how can we > estimate the autoregressive parameter of beta (pi8) and get its standard > error? How can we relax the random walk version of beta and estimate pi8, > instead of assigning a value of 1.0 or 0.0 to it as in the TVARYING.PRG? > > Thanks for any suggestions. See the example on page 14-96 that uses the command "FIND." However FIND does not give the standard error of the estimate. I used a similar program, and after wandering around a little the estimate seemed to settle at the initial value or half of the initial value (I tried to estimate PI7). Might well be a bug in my program. Did anybody else encounter a similar problem? I would recommend the command "MAXIMIZE." CAN ERSOZ UNIV. OF TEXAS AT DALLAS ---------- End of message ---------- From: Xavier Rousseau <100530.3144@compuserve.com> To: "RATS Discussion List" Subject: Estimation of structural models Date: 24 Mar 96 04:21:49 EST Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Mercury MTS v1.21 Does anyone know how to estimate a structural model such as y(t) = y(t-1) + a(t) + e(t) ; e(t) is white noise a(t) = a(t-1) + n(t) ; n(t) is white noise and uncorrelated with e(t) Random walk with stochatic drift ? The problem is to estimate var(e(t)), var(n(t)), and to get at each t an estimation for a(t). Of course the model can be written in a state space form with y(t) = ( y(t-1), 1) * (1,a(t))' + e(t) as the mesurement equation and (1,a(t)) = (1,a(t-1)) + (0,n(t)) as the state vector The kalman filter appears to be a powerful and convenient method of estimation. But how can we initiate the processus in Rats with the Kalman instruction as the variances are unknown at the beginning. Thanks for help X Rousseau Modem, Paris X ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Estimation of structural models -Reply Date: Mon, 25 Mar 1996 09:15:22 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Estimation of structural models -Reply -Reply Date: Mon, 25 Mar 1996 11:01:41 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Rob Trevor To: "RATS Discussion List" Subject: Re: Estimation of structural models Date: Mon, 25 Mar 1996 12:05:25 +1100 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Mercury MTS v1.21 Xavier > Does anyone know how to estimate a structural model such as > > y(t) = y(t-1) + a(t) + e(t) ; e(t) is white noise > a(t) = a(t-1) + n(t) ; n(t) is white noise and uncorrelated with e(t) > Random walk with stochatic drift > É There may well be a few ways to do this in RATS, but one which immediately springs to mind is to use the MAXIMISE(INITIAL=xxx,É) instruction. You'd probably want to make a distributional assumption about e(t) and n(t), and then maximise the likelihood function. You may also wish to look at the GARCH stuff in the manual to see how to code a recursive model. Cheers Rob ----------------------------------------------------------------------- Rob Trevor Associate Professor Centre for Studies in Money, Banking and Finance Macquarie University, NSW, AUSTRALIA 2109 Internet: robt@efs.mq.edu.au VOICE: +61 (2) 850-8539 FAX: +61 (2) 850-7281 or +61 (2) 418-1847 ----------------------------------------------------------------------- ---------- End of message ---------- From: Rob Trevor To: "RATS Discussion List" Subject: Vacation Mail for Tim Ng Date: Mon, 25 Mar 1996 12:21:33 +1100 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Mercury MTS v1.21 I've just sent the appended message to the RBNZ. If you are subscribed to *any* email list, and wish to use a 'vacation mail' type facility PLEASE either (a) ensure that this facility is smart enough to post no more than one notice to any address, or (b) none to email lists, or (c) unsubscribe from the lists while you are away. Please consider your colleagues email boxes! Thanks Rob ------------------------------ Dear Post Master One of your users (Tim Ng ) appears to have set up a "vacation mail" facility. Unfortunately, this facility doesn't seem too smart and is responding to EVERY message on the "RATS Discussion List" . I'd appreciate it if you could take steps to prevent this list from receiving these vacation notices. Otherwise I will have to remove your user from this list. At 11:01 AM 25/3/96, Tim Ng typed: > I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact > details. Many thanks Rob Trevor RATS-L owner ----------------------------------------------------------------------- Rob Trevor Associate Professor Centre for Studies in Money, Banking and Finance Macquarie University, NSW, AUSTRALIA 2109 Internet: robt@efs.mq.edu.au VOICE: +61 (2) 850-8539 FAX: +61 (2) 850-7281 or +61 (2) 418-1847 ----------------------------------------------------------------------- ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Re: Estimation of structural models -Reply Date: Mon, 25 Mar 1996 13:17:26 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Estimation of structural models -Reply -Reply -Reply Date: Mon, 25 Mar 1996 13:17:24 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Vacation Mail for Tim Ng -Reply Date: Mon, 25 Mar 1996 14:01:29 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Re: Estimation of structural models -Reply -Reply Date: Mon, 25 Mar 1996 14:17:29 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Estimation of structural models -Reply -Reply -Reply -Reply Date: Mon, 25 Mar 1996 14:17:31 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Vacation Mail for Tim Ng -Reply -Reply Date: Mon, 25 Mar 1996 17:19:41 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Re: Estimation of structural models -Reply -Reply -Reply Date: Mon, 25 Mar 1996 18:26:31 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Estimation of structural models -Reply -Reply -Reply -Reply Date: Mon, 25 Mar 1996 18:25:32 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Vacation Mail for Tim Ng -Reply -Reply Date: Mon, 25 Mar 1996 19:33:38 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Re: Estimation of structural models -Reply -Reply -Reply Date: Mon, 25 Mar 1996 20:16:44 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Estimation of structural models -Reply -Reply -Reply -Re Date: Mon, 25 Mar 1996 20:16:44 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Vacation Mail for Tim Ng -Reply -Reply Date: Mon, 25 Mar 1996 21:00:45 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) -Reply I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Re: Estimation of structural models -Reply -Reply -Reply Date: Mon, 25 Mar 1996 21:00:46 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) -Reply I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Tim Ng To: "RATS Discussion List" Subject: Estimation of structural models -Reply -Reply -Reply Date: Mon, 25 Mar 1996 21:00:47 +1200 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: X-Mailer: Novell GroupWise 4.1 (via Mercury MTS v1.21) -Re -Reply I am away from the Bank until Wed, April 3. Debbie May x3891 has my contact details. ---------- End of message ---------- From: Rob Trevor To: "RATS Discussion List" Subject: Tim Ng's Mini Mail Storm! Date: Mon, 25 Mar 1996 21:40:07 +1100 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Mercury MTS v1.21 OK folks, the mail storm we've just been experienced has forced me to act immediately. I've just attempted to delete Tim Ng from the list by remote control (I'm at home at present). If that fails for some reason, I will do it manually when I get to work in the morning. Tim - you're most welcome to resub when you get back, but please be more careful with automated mail features next time you take a break! I simply had to delete you because you've managed to create a mail loop. What seems to be happening is that each time a list message arrives at Tim's mailbox, the vacation mail program at the RBNZ is sending the vacation message TO THE LIST ADDRESS. Since Tim is a member of the list, the list processor then copies Tim's own vacation message back to Tim, which generates another vacation mail message to the list, whichÉyou get the idea. The net result is that for EACH genuine RATS-L message posted to the list while Tim's away, we will all get a v. large number of vacation mail responses - the number will be limited purely by the speed of the RBNZ <-> Macq Univ mail link and mailer cycle times. I hope that we can try to avoid this in the future. Cheers Rob Trevor RATS-L List owner ----------------------------------------------------------------------- Rob Trevor Associate Professor Centre for Studies in Money, Banking and Finance Macquarie University, NSW, AUSTRALIA 2109 Internet: robt@efs.mq.edu.au VOICE: +61 (2) 850-8539 FAX: +61 (2) 850-7281 or +61 (2) 418-1847 ----------------------------------------------------------------------- ---------- End of message ---------- From: Rob Trevor To: "RATS Discussion List" Subject: Tim Ng's mail Date: Tue, 26 Mar 1996 10:19:12 +1100 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Mailer: Mercury MTS v1.21 The following just arrived from RBNZ. Many thanks to Harvey Watts for his prompt actions. Let this be a lesson to all of us on the list on the downside of untested automated email response systems. I will now try to re-sub Tim so that he doesn't miss out on the real action of the list! Rob RATS-L List Owner << start of forwarded material >> >Date: Tue, 26 Mar 1996 09:31:12 +1200 >From: Harvey Watts >Subject: Tim Ng's mail > >Rob, > >Our apologies for the inconvenience caused by Tim's mail. I >have managed to get into his mail account and disable the >reply rule, so no further mail should be generated. > >We will certainly give consideration as to how we get our >staff to set up rules in the future. > >Harvey Watts >Computer Services >Reserve Bank of NZ. > << end of forwarded material >> ---------- End of message ---------- From: "F.Y.Kumah" To: "RATS Discussion List" Subject: FEVD IN SVARs Date: Thu, 28 Mar 1996 09:39:59 MET Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: Organization: Tilburg University X-mailer: Pegasus Mail for Windows (v2.23) (via Mercury MTS v1.21) Hello and G'day, Can someone out there give me a handy program for estimating the forecast error variance decompositions (and thier standard errors) for Structural VARs? I have programmed my own but the standard errors tend to be rather too large to be plausible. And I have tried using Giannini's SVAR.SRC and VMA.SRC but the program breaks down midway and reports that the model (which is clearly identified) is unidentified. Best Regards Francis **~~**~~**~~**~~**~~**~~**~~**~~**~~**~~**~~**~~****~~**~~**~~**~~** Francis Y. Kumah CentER for Economic Research P. O. Box 90153 5000 LE Tilburg The Netherlands Phone (+31) 13 - 4668 221 Fax (+31) 13 - 4663 066 My Home Page: http://cwis.kub.nl/~few5/center/phd_stud/kumah/home.HTM **~~**~~**~~**~~**~~**~~**~~**~~**~~**~~**~~**~~****~~**~~**~~**~~** ---------- End of message ---------- From: Sukudhew Singh To: "RATS Discussion List" Subject: Problem with Nonlinear Equation System Date: Sun, 31 Mar 1996 02:14:57 -0600 Errors-to: Reply-to: "RATS Discussion List" Sender: Maiser@efs1.efs.mq.edu.au X-listname: MIME-version: 1.0 X-Mailer: Windows Eudora Light Version 1.5.4b13 (32) (via Mercury MTS v1.21) Content-type: text/plain; charset="us-ascii" Content-transfer-encoding: 7BIT Hi, I have been trying to estimate a system of nonlinear equations without much success. I keep getting the error message: ## SR10. Missing Values And/Or SMPL Options Leave No Usable Data Points This is the first time I have ever estimated a system of nonlinear equations in RATS. I ran the sample program CONSUMER.PRG (which comes bundled with RATS) and it worked just fine. I also know for sure that the system I am working with has been successfully estimated using the SYSNLIN procedure in SAS, so I must be missing something in my RATS program. Even doubling the number of observations in the data set does not solve the problem. Many hours with the RATS manual and browsing through the help file have gotten me nowhere. I am hoping that someone on this group with experience in estimating such systems would be able point out what I might be doing wrong. I attach both the program and the data I am using. I would appreciate any helpful suggestions. Thanks. Sukhdave S. Gill singhs@ctrvax.vanderbilt.edu ************** ** PROGRAM ** ************** CALENDER 1965 ALLOCATE 1989:1 OPEN DATA I:\S.RAT DATA(FORMAT=RATS) * free parameters NONLIN A0 A1 A2 A3 A4 B0 B1 B2 B3 L * equations FRML DD Q = A0 + A1*P + A2*Y + A3*P*Z + A4*Z FRML SS P = L*((-1*Q)/(A1 + A3*Z))+B0 + B1*Q + B2*W1 + B3*W2 * initial values COMPUTE A0=A1=A2=A3=A4=0.0 COMPUTE B0=B1=B2=B3=L=0.0 * estimate system NLSYSTEM / DD SS *********** ** DATA ** *********** Q P Y Z W1 W2 K C D 24832 0.0453 57523 4.0 0.0231 0.00433 0.3082 1082 0 26856 0.0487 64388 5.0 0.0260 0.00480 0.3412 1253 0 29422 0.0503 69064 4.6 0.0277 0.00520 0.3535 1440 0 33718 0.0575 75418 6.3 0.0334 0.00568 0.3724 1828 0 38854 0.0677 83026 7.2 0.0422 0.00614 0.4045 2458 0 44362 0.0734 89116 6.0 0.0504 0.00666 0.4217 3063 0 49944 0.0641 97290 3.6 0.0407 0.00706 0.4086 2983 0 58400 0.0597 108629 3.6 0.0361 0.00724 0.4002 3200 0 71286 0.0667 127372 5.5 0.0442 0.00819 0.4264 4423 0 88776 0.8520 152111 7.8 0.0655 0.00938 0.4359 7229 0 103556 0.0821 171540 7.4 0.0592 0.01102 0.4458 7964 0 119813 0.0797 197924 8.9 0.0573 0.01232 0.4483 9110 0 141745 0.0757 217879 7.3 0.0540 0.01320 0.4547 10323 0 172155 0.0784 241604 8.7 0.0584 0.01449 0.4366 12971 0 213128 0.0950 276096 11.7 0.0816 0.01678 0.4293 20117 0 260828 0.1091 309891 12.8 0.1011 0.01948 0.4325 28481 0 329782 0.1361 355994 17.7 0.1310 0.02223 0.4581 43605 1 367878 0.1382 374442 13.7 0.1311 0.02718 0.3723 49851 1 367409 0.1050 415717 9.3 0.0891 0.02867 0.3787 36381 1 387517 0.1053 444735 11.1 0.0909 0.03004 0.3498 38912 1 425864 0.0968 447988 9.4 0.0814 0.02990 0.3579 38963 1 451578 0.0888 514631 9.0 0.0726 0.03175 0.3737 37673 1 469781 0.0828 551334 8.1 0.0666 0.03387 0.3745 36636 1 490521 0.0889 601508 9.5 0.0725 0.03720 0.3933 40732 1 527614 0.1082 648537 12.1 0.0909 0.04099 0.4053 51785 1 ---------- End of message ----------