1. NAME AND TITLE

LSMOD/GLSMOD: A Least-Squares Computational "Tool Kit".

2. CONTRIBUTOR

Argonne National Laboratory, Argonne, Illinois.

3. CODING LANGUAGE AND COMPUTER

Microsoft Fortran 3.0; IBM PC's and compatibles.

4. NATURE OF PROBLEM SOLVED

LSMOD and GLSMOD can be used in a very general way to solve simple and generalized least-squares problems. The routines can handle input data sets with correlated errors and produces output that includes errors and correlations for solution parameters. Special applications using the routines require tailoring of the input/output routines and/or iterative control (for LSMOD).

5. METHOD OF SOLUTION

Matrix formalisms of well-known least-squares algorithms are used and are described in the documentation.

6. RESTRICTIONS OR LIMITATIONS

For increased precision, it is recommended by the author that REAL variables be assigned 8 bytes rather than 4 bytes.

7. TYPICAL RUNNING TIME

Execution times for LSMOD and GLSMOD using the sample problems on a Northgate 486/66 were approximately one second.

8. COMPUTER HARDWARE REQUIREMENTS

An IBM PC or compatible running MS DOS 3.1 or higher. A hard disk is not required.

9. COMPUTER SOFTWARE REQUIREMENTS

MS Fortran 3.0 or higher is recommended. LSMOD and GLSMOD were tested at RSIC using the included sample input files on a Northgate 486/66 running MS-DOS 6.2 and using MS Fortran 5.0.

10. REFERENCE

D. L. Smith, "A Least-Squares Computational 'Tool Kit'," ANL/NDM-128 (April 1993).

11. CONTENTS OF CODE PACKAGE

The referenced document and one 1.2 MB diskette in DOS format are included. The code package contains sample input and output, the author's executables, and the LSMOD and GLSMOD source files.

12. DATE OF ABSTRACT

January 1994.