8/27/2001
This is Fortran archive package of Non-metric Multidimensional
Scaling Method (NMDS) developed by Y-h. Taguchi and Y. Oono.
Ver. 1.2.
For installation, see INSTALL.
For instruction, see INSTRUCTION.
For differences between versions, see CHANGELOG
The construction of this package is as follows:
MDS_final : directory includes Fortran source codes
MDS_final/utilities : directory includes utilities
MDS_final/distcal1 : directory includes example of distcal1.f
MDS_final/doc : some technical details
Contents of MDS_final :
README : This file.
INSTALL : installation how to.
INSTRUCTION : brief instruction of this package.
Makefile : is Makefile
mds.f : main routine.
mds-cgi.f : main routine for output of geomview format
Subroutines
distcal.f : compute distance matrix of obtained configuration.
distcal1.f : compute distance matrix of target configuration.
(In this package, it reads d dimensional vector from file fort.51.
Modify it as you like) It is same as distcal1/distcal1_vec.f.
fcal.f : compute 'Force' along bonds.
init.f : initialization.
prep.f : compute rank order data for targeted configuration.
renorm.f : centered and normalize vector.
test.f : compute significant level for obtained configuration.
geomview_out.f : output geomview format of obtained config.
rand.f: uniform random number generator form SLATEC (*)
spsort.f : quick sort routine form SLATEC (*)
Subroutines called by spsort.f itself or those called by spsort.f is
as follows:
fdump.f :
i1mach.f :
j4save.f :
xerhlt.f :
xercnt.f :
xermsg.f :
xerprn.f :
xersve.f :
xgetua.f :
(*)SLATEC
http://phase.etl.co.jp/mirrors/netlib/slatec/
Contents of MDS_final/utilities:
Makefile : is Makefile
ranvec.f : generate vectors whose components are random numbers
hypercube.f : generate hypercube (3 dimensional)
Contents of MDS_final/distcal1 :
distcal1_vec.f : compute distance matrix of target configuration,
d dimensional vector from file fort.51.
Content of MDS_final/doc :
README : some explanation
paper.pdf : much more technical details (unpublished)
This package is tested on Kondara 2.0 (http://www.kondara.org),
which is one of major distribution of Linux.
It employs usual g77 compilers, thus this package would be
executed for any computer system with Fortran compiler.