arg@ccvr1.ncsu.edu (Ron Gallant) (12/23/90)
nlmdl is a C++ implementation of the statistical methods in A. Ronald Gallant, "Nonlinear Statistical Models," New York: John Wiley and Sons, 1987, ISBN 0-471-80260-3, using a matrix class realmat that is distributed with it. The matrix class may be of some independent interest. The program computes least squares estimates for a univariate nonlinear regression model and generalized least squares estimates for a multivariate nonlinear regression model. Use the SUR option for both. Other options can be set to correct variance estimates for heteroskedasticity and/or serial correlation. For simultaneous nonlinear systems, possibly in implicit form, it computes three-stage least-squares estimates, TSLS option, and for nonlinear dynamic systems, possibly in implicit form, generalized method-of-moments estimates, GMM option. The documentation consists of this file, the two header files nlmdl.h and realmat.h, and the examples in the directories ch1eg1, ch5eg1, ch6eg1, and ch6eg2. As distributed, the code is set up for use with GNU's g++ compiler on a Unix box. To check using one of the examples, e.g. ch1eg1, cd to ch1eg1 and type make -f ../Makefile. Each directory has a file detail.bak created using nlmdl > detail.dat that you can use for comparison. Although intended for use with GNU's g++ compiler, the code can be used with Borland's Turbo C++ compiler. To use Turbo C++, permute the #undef and #define for compiler choice in usual.h, copy the contents of, e.g., ch1eg1 to the working directory, and list every file with a .cc extension in the .PRJ file except nlopr.cc. Small models, such as ch1eg1, execute correctly when compiled with the large memory option but larger ones, such as ch5eg2, run out of memory during execution. I haven't tried Turbo's overlay and memory expansion features to see if larger models can be accommodated. The code hasn't been tested with other C++ compilers. Copyright (C) 1990 by A. Ronald Gallant. Permission to use, copy, modify, and distribute this software and its documentation for any purpose and without fee is hereby granted, provided that the above copyright notice appear in all copies and that both that copyright notice and this permission notice appear in supporting documentation. This software is provided "as is" without any expressed or implied warranty. ------------------------------------------------------------------------------- Professor A. Ronald Gallant Phone: 919-737-2531 Department of Statistics FAX: 919-737-7591 North Carolina State University e-mail: arg@ccvr1.cc.ncsu.edu Campus Box 8203 ftp: ccvr1.cc.ncsu.edu (128.109.212.20), Raleigh NC 27695-8203 USA user: anonymous, directory: pub/arg -------------------------------------------------------------------------------