home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Usenet 1994 January
/
usenetsourcesnewsgroupsinfomagicjanuary1994.iso
/
sources
/
reviewed
/
volume01
/
ols
/
part00
next >
Wrap
Internet Message Format
|
1991-11-06
|
4KB
From: Ajay Shah <ajayshah%monty@rand.org>
Subject: v01i035: ols - linear regression tool, Part00/04
Newsgroups: comp.sources.reviewed
Approved: csr@calvin.doc.ca
Submitted-by: Ajay Shah <ajayshah%monty@rand.org>
Posting-number: Volume 1, Issue 35
Archive-name: ols/part00
Environment: Unix (requires ANSI C)
Note: this program was designed on SUNs using gcc. It is a
useful program and is somewhat portable, although getting it
to work in other environments may take some work.
Tested Environments:
DECstation 2100, ULTRIX 4.1
Decstation 3100, ULTRIX, vcc (VaxC-compiler).
VAXstation 3200, Ultrix, gcc (version 1.39)
Dynix, gcc (version 1.39)
Sun, gcc (-ansi)
VAX/VMS, VAX C-Compiler
IBM PS2/70, OS2 v.1.1, Microsoft C-Compiler v6.00A (WILL NOT RUN)
AT&T 3B1 (gcc) (WILL NOT BUILD)
SCO UNIX 3.2.2 using gcc 1.43 (WILL NOT RUN)
Dates:
Submission Received: June 19, 1991
Reviews Returned: July 15, 1991
Revised Submission Received: September 13, 1991
Reviews Returned: October 6, 1991
Revised Submission Received: November 5, 1991
Author's Summary:
-----------------
This is a Unix tool to do linear regressions (OLS). It is fast, works
without much fuss and integrates well into Unix. It does many of the
things a normal statistics package does by way of linear regressions.
...
Some of the numerical-analysis C source in this package is not commonly
available in the public domain. It should be useful in it's own right.
All the code is clean, ANSI C. All IO is StdIn, StdOut and StdErr. It
should port easily to other systems -- though I have never tried
anything except for Suns.
...
EPP: a utility for typesetting results of statistical estimations
This is a related utility shipped with ols. It is actually a general
tool for typesetting results of statistical estimation into LaTeX form;
it is not limited to use with OLS. It is in the file epp2tex.shar. If
you want this functionality, then unpack this sharfile and read the
README inside it.
This is strictly optional! If you have no interest in typesetting
results of OLS estimation into LaTeX form, you do not need to look at
epp2tex.shar.
Reviewers' Comments:
--------------------
I think this package is useful. Somehow I hope I had this when I
struggled in the physics laboratory and tried to fit my observations
with a ruler. The interface is simple and neat, there is a very
limited set of features and the features are, as far as I can
understand, well chosen.
...
I would recommend this package for use 'as is' for anyone who
understands Linear Regression methods. For someone not very familiar
with the methods, I would not.
...
As for the source, I think I saw somewhere a sentence 'This is my first
program in C' [paraphrase] by the author. The code is very readable,
which is, if this is his/hers first C-program, a small miracle.
---------------------------------------------------------------------------
OLS seems to behave well on a SUN, if compiled with gcc. Otherwise be
prepared to hack it. I recommend it to be posted; however, I also
recommend a note about 'possible minor problems when porting to other
environments than Sun/SunOs/gcc' included in the posting.
---------------------------------------------------------------------------
`ols' is a package for doing linear regression. As such, it can be a
very useful package and I recommend it for use to anyone who needs it.
My only problem with the current implementation is that the portability
of the code is not as good as one often sees in programs posted to the
net; expert assistance may be needed for installation. The program
comes with examples to help you get started, and all the proper
documentation, as well as EPP, a package for printing the results in
LaTeX.
exit 0 # Just in case...