Elsevier Science Home
Computer Physics Communications Program Library
Full text online from Science Direct
Programs in Physics & Physical Chemistry
CPC Home

[Licence| Download | New Version Template] abcg_v1_0.gz(17 Kbytes)
Manuscript Title: Search program for significant variables.
Authors: M.J. O'Connell
Program title: LINCOM
Catalogue identifier: ABCG_v1_0
Distribution format: gz
Journal reference: Comput. Phys. Commun. 8(1974)49
Programming language: Fortran.
Computer: IBM 360/195.
Operating system: IBM/OS, HASP.
RAM: 13K words
Word size: 32
Peripherals: magnetic tape.
Keywords: General purpose, Nuclear physics, High energy, Significance search, Dispersion matrix, Significant variables, Eigenvalues, Eigenvectors, General experiment.
Classification: 4.9, 17.4.

Nature of problem:
The coordinates associated with readings of one or more independent variables are converted into significant linear combinations of the original variables.

Solution method:
Eigenvalues and eigenvectors of the 'dispersion' matrix for the problem are calculated, and the 'position-vector' of each observation is analyzed into the components along the normalized eigenvectors. If the eigenvalues are arranged in decreasing order of the linear combination in corresponding order, the most important linear combination is the first and the least important is the last. If some eigenvalues are very small relative to the largest, then the corresponding linear combination may be ignored or set to zero. In this way the complexity of a problem can be reduced. The derivation of momenta from spark chamber coordinates using this method has been discussed by Wind, from whose original program the present version has been standardised.

Unusual features:
Complete data set of observations is not held in core since the dispersion matrix is recomputed after each record is read.

Running time:
5 s on IBM 360/195 (including 2.6 s to compile) for analysis of 500 events with 8 independent coordinates.