Having problems with message search?
I am looking for any information on how to calculate the standard error of a predicted score from a multiple regression equation. This SE allows you to calculate a confidence interval about a predicted score.
For simple linear regression, the equation is:
SEp = SEE * sqrt(1 + (1/N) + ((X-Xbar)^2/sum(Xi - Xbar)^2))
where SEp = standard error of predicted score, SEE = standard error of estimate for the regression equation, X = score on the predictor, and Xbar = mean of the predictor. The CI then is Y' + t * SEp, and t is the t value for the specific alpha and df.
My copy of Pedhazur states that when you have multiple predictors, the algebraic formulas become "unwieldy", so he presents the calculation using matrix algebra. I'd like to write a little program to calculate this and it would be easier to deal with unwieldy algebra than to figure out how to code matrix algebra. So does anybody have this calculation tucked away some where that is in an algebraic form?
Thanks in advance.
Joe
 | ||
| Joseph P. Weir, Ph.D. FACSM
Professor and Research Coordinator
Division of Physical Therapy
Des Moines University - Osteopathic Medical Center 3200 Grand AvenueDes Moines, IA 50312 515.271.1733 FAX 515.271.1714 joseph.weir@... www.dmu.edu | ||
Send Email
