Re: Need multiple regression source/book

If you are doing this just for curiosity, then re-inventing the wheel will do no harm.

If this is for forensic or even management decision making, where there are potentially serious consequences in people's lives. I strongly suggest that you use a standard, well-established statistical package.

Excel is a spread***. Use it as a spread***. Use a statistical package to do statistics. I would not testify in court or advise on other serious decisions based on ad-hoc software.

to see the algorithms involved
go to www.spss.com
click <login to support>
login as "guest" password "guest"  click "ok"
click <statistics>
click <algorithms>
click . . .

You say the system "cannot" rely on third party software. Did some manager set this specification. Is (s)he setting you up as the fall guy in a lawsuit? Or is someone trying to delay the provision of a statistical analysis.

As a really good book on regression check this cut-and-paste from the Library of Congress Catalog.
Main Title: Applied multiple regression/correlation analysis for the behavioral sciences / Jacob Cohen ... [et al.].
Edition Information: 3rd ed.
Published/Created: Mahwah, N.J. : L. Erlbaum Associates, 2003.
Related Names: Cohen, Jacob, 1923-
Cohen, Jacob, 1923- Applied multiple regression/correlation analysis for the behavioral sciences.
Description: xxviii, 703 p. : ill. ; 26 cm. + 1 CD-ROM (4 3/4 in.)
ISBN: 0805822232 (hard cover : alk. paper)
Notes: Rev. ed. of: Applied multiple regression/correlation analysis for the behavioral sciences / Jacob Cohen, Patricia Cohen. 2nd ed. 1983.
The CD-ROM contains the data for almost all examples as well as the command codes for each of the major statistical packages for the tabular and other findings in the book.
Includes bibliographical references (p. 655-669) and indexes.
Subjects: Regression analysis.
Correlation (Statistics)
Social sciences--Statistical methods.
LC Classification: HA31.3 .A67 2003
Dewey Class No.: 519.5/36 21

Even if it is not for federal courts this is a very good standard for scientific evidence.

Main Title: Reference manual on scientific evidence.
Edition Information: 2nd ed.
Published/Created: Washington, DC (One Columbus Circle, N.E., Washington, DC 20002-8003) : Federal Judicial Center, 2000.
Related Names: Federal Judicial Center.
Description: vii, 639 p. : ill. ; 26 cm.
Notes: Includes bibliographical references and index.
Subjects: Evidence, Expert--United States.
LC Classification: KF8961 .R44 2000
Dewey Class No.: 347.73/67 21
Govt. Doc. No.: JU 13.8:SCI2
Geog. Area Code: n-us---
GPO Item No.: 743-C-2

CALL NUMBER: KF8961 .R44 2000

You can download the section on statistical evidence at
http://www.fjc.gov/public/home.nsf/autoframe?openform&url_r=pages/556&url_l=index
You can download the chapter on multiple regression at
http://www.fjc.gov/public/home.nsf/autoframe?openform&url_r=pages/556&url_l=index


Hope this helps.

Art
Art@xxxxxxxxxxxxx
Social Research Consultants
University Park, MD  USA
(301) 864-5570


Tim Witort wrote: 
I am writing a computer program that must perform a multiple
linear regression analysis to measure the influence of various
employee attributes on salary.  The basic goal is to determine
whether race or sex is a statistically significant factor in
comparison to a couple of other factors such as time in job,
education, performance level, etc.

Excel can accomplish this with its regression tool - giving
a t-stat for the various variables that are used, however
this system cannot depend on Excel or other third-party
tools.

I have a good grasp of general statistics.  Can anyone recommend
a source or book with actual computations and formulas that
would allow me to code this?  Thus far, my web searches have
found only descriptions of the theories behind multiple regression,
not the actual math that accomplishes the anlaysis.  And I've
found next to nothing on calculating t-stats for the variables
following the regression.

Any recommendations would be appreciated.

-- TRW
_______________________________________
t r w 7
at
i x  dot  n e t c o m  dot  c o m
_______________________________________

.