multiple regressions with mixed logic and real parameters

[agolkar]

Hi,
I am looking for references to guide me through multiple linear/nonlinear regressions based on mixed logic and real parameters. I have to develop CERs for a project in which some design variables are boolean (option A is ON / option A is OFF), whereas other variables are real (i.e. speed, mass, and so forth). I have generated the data I need that now I have to fit, and I am looking for references or texts that address this specifical problem.

Thanks!

 

[dogtop]

Have you inquired in the "Statistics in engineering" forum?
Regulars there might be able to help.

 

[katmar]

This information was conveyed to me by a non-member in a private email, so I cannot vouch for it or offer any further information.

There is a book by N.R. Draper and H. Smith, Applied Regression Analysis (J. Wiley, ISBN 0-471-029995-5), where in chap.5, paragraph 5.4 there is a description of the use of "dummy" variables - factors that have two or more distinct levels.


Katmar Software
Engineering & Risk Analysis Software

http://katmarsoftware.com

 

[PredictorX]

Numeric dummy variables can be used to represent categories as 0/1 flags. Note that you will need one dummy variable less than the total number of categories:

0 male
1 female

or...

0 1 red
1 0 green
0 0 blue


-Will Dwinnell
Data Mining in MATLAB