NED@YMIR.BITNET (Ned Freed) (09/28/87)
In response to several requests I recently posted some information about MATHLIB on the info-pmdf mailing list. This has generated a suprising amount of mail, including mention of the fact that a couple of queries about MATHLIB had appeared on info-vax a while back. Well, I must have missed them somehow. How embarassing. Sorry about the LONG delay in responding to them. Anyhow, here is some background information on the MATHLIB package: Most people are familiar with the concept of a statistical analysis package. There are lots of them available: SPSS, BMDP, Minitab, SAS, GLIM, Systat, TSP, STP, and many others. The idea embodied by these packages is simple. They provide "canned" statistical procedures for application to the user's data. Such programs make it possible to perform very sophisticated analyses without actually doing any real programming; the user need only learn the commands relevant to the type of analysis he or she wants to perform. This is MUCH easier than trying to write a program to do the same thing, even if you happen to have a subroutine library available that will do the actual analysis. MATHLIB is a package similar in form to a statistics package, except that it performs various types of mathematical modelling and analysis and not statistical analysis. MATHLIB can accomodate models expressed in many different ways, including nonlinear equations, Fourier transforms, ordinary differential equations, partial differential equations, integral equations, and difference equations. Models are entered as mathematical expressions via an extremely user-friendly interactive command interface which includes a convenient symbolic manipulation facility. Models can be fit to data using nonlinear least squares fitting techniques. Smoothing and interpolation capabilities are also provided. MATHLIB easily handles complex models and large data sets. Facilities are provided to store, retrieve, transform, analyze, and fit models to data. A key component of MATHLIB is its ability to produce graphics output at any time. MATHLIB can produce: (1) Line plots and scattergrams, with or without error bars. (2) Plain, grouped and stacked bar charts and histograms. (3) Pie charts. (4) Contour plots. (5) Three dimensional surface plots. (6) Vector (flow) field diagrams. (7) User-defined plot procedures (using a built-in graphics language). MATHLIB graphics output is device-independent. If your device is supported (and over 150 graphics devices are currently supported), you can take advantage of all the graphics capabilities of MATHLIB. Even if you do not have any supported output devices, MATHLIB provides a rudimentary character graphics facility that can be used on virtually any output device. MATHLIB also provides various utilities useful for system performance monitoring and management on VAX/VMS systems. In particular, MATHLIB can read and analyze data output by the VMS MONITOR facility. MATHLIB is well-suited to this task since analysis of system performance data is essentially mathematical in nature, and presentation of such data is usually done best in a graphical format. MATHLIB also provides an extensive library of user-callable routines so that users may develop modelling software of their own based on MATHLIB's symbolic manipulation facilities. MATHLIB is well documented. Complete online help is available along with a comprehensive reference manual (1000 pages), a user's guide and a variety of application notes describing specific uses of MATHLIB. Ongoing support and maintenance for MATHLIB is available. A new version of MATHLIB is released approximately every year. Maintenance patches and minor revisions are released as they are needed. MATHLIB is currently available for VAX systems running VMS. A version for the Sun should be released within the next few months. MATHLIB consists of approximately 80,000 lines of Pascal implementing the driver programs, command parser, and symbolic manipulation facilities, 80,000 lines of FORTRAN implementing the numerical analysis routines, and about 100,000 lines of FORTRAN implementing the graphics facilities. Source is not included in the regular distribution but is available by special arrangement for development purposes. If you want additional information about MATHLIB please send me mail and include a snail mail address. I'll see that some literature gets sent your way. Ned Freed (ned@ymir.bitnet) MATHLIB Project Department of Mathematics Harvey Mudd College Claremont, CA 91711 (714) 621-8006