Houston H. Stokes PhD.
1700 E 56th Street # 3001
Chicago, IL 60637
(773)-643-4384
//www.uic.edu/~hhstokes/hhs1.html
Mr. Eric Valentine
Publisher
Quorum Books
88 Post Road West
P. O. Box 5007
Westport, CT 06881-9990
Dear Eric:
I want to update you on my proposed plans for the future in a number of areas. I have been very pleased with my student's response the book Explorations of New Methods of Financial Modeling which I did with Hugh Neuburger. I used this book in the advanced Time Series Classand, based on the response, will do so again. This book does not required revision since it is completed research. I plan more such books in the future.
I appreciate the job Quorum has done on the two editions of my book Specifying and Diagnostically Testing Econometric Models which were published in 1991 and 1997. As you remember my books are supplied 100% camera ready. They are edited by my wife Diana, who is a professional editor. There is no type setting of any kind by Quorum. In the past I had to develop specialized macros for math type setting in Word Perfect 5.1. The books were not set in proportional type. Recently I have shifted into WORD97 and am now able to have the book 100% Times Roman except for computer output and command files. I hope to get a third edition ready for 1997+6 = 2003. I hope this is OK with you. My book documents the b34s® program and is used in 3 of my classes and will be used as long as I am teaching. I enclose two chapters of the "revised" format so you can see what I am talking about and see the improvement in appearance.
In the last two years I have been involved in a major software development effort. I have built a 4th generation mathematical programming language into b34s that, among other things, allows estimation of a whole class of nonlinear models that were not possible to be estimated in the past. The b34s MATRIX command is a true programming language like Matlab® and Gauss® but wityh specialized commands focused on econometric research. While it might be possible to expand the Nonlinear Chapter in Specifying and Diagnostically Testing Econometric Models to over 150 pages, it might be better to have a distinct book dealing with the topic of nonlinear modeling with many examples. That is what I am proposing to do. The "working title" would be Applied Nonlinear Modeling for Finance and Economics Using a 4th Generation Language. Nonlinear modeling is very important in Finance and Econometrics. The models estimated and discussed in Explorations of New Methods of Financial Modeling were for the most part nonlinear. The problem was that the software available (RATS) did not allow monitoring of the solution progress or estimation of anything but a very restrictive class of maximum likelihood models that involved summing log terms. My proposed new book will illustrate nonlinear modeling of a number of data sets supplied by the Bureau of Standards as well as a number of economic and finance models. Software to perform the calculations will be available on both Windows 95/98/NT platforms and Linux as well as Sun and RS/6000 unix machines. My new software allows the researcher to see the solution proceeding in n dimensions.
An alternative to this moe focused book, is longer work that illustrates other calculations that are possible with the new software. This might go into a future edition of the proposed book.
I will try to illustrate the advance that has been made. Most if not all nonlinear estimation systems either require the user to either:
Program a complex function in Fortran and estimate the model using a subroutine library.
Use a higher level programming system whereby a restricted form of the model is estimated.
Write the model completely in a higher level language.
The first approach is fast but requires extensive programming knowledge. The second approach is a compromise. Speed is gained at a loss of generality of the models that are possible since the structure of the model is hard wired in the software. With this approach the user does not have control of the process. The completely higher language implementation suffers a speed loss. The new system I have developed is a hybrid in that it allows the user to program the function in a higher language (in place of Fortran) yet still link with a Fortran solution subroutine such as that in the IMSL (International Mathematical Subroutine Library). General models can be estimated without major speed loss. In addition the solution can be graphically monitored, something not possible with other systems.
I am not ready to sign a contract but am interested in what you think of the proposed project. I can assure you I will do a very good job. As you know I like to work way ahead. I do not like last minute efforts.
Sincerely,
Houston H. Stokes
Rough Outline of
Applied Nonlinear Modeling for Finance and Economics Using a 4th Generation Language
Houston H. Stokes
Department of Economics
University of Illinous
Preface
Introduction
Nonlinear least squares and Maximum Likelihood Estimation Theory
Programming Nonlinear least squares Software using a 4th Generation Language
Nonlinear Least Squares Models using the Levenberg-Marquardt Algorithm
Nonlinear Maximization
Estimating ARCH, GARCH and GARCH-M Models
Conclusion