Re: Prediction Model Question (long)

[David A. Heiser]

>From Burrill and Ulrich's discussion.

All this orthagonalization is fine. To me the bottom line is still the
residuals and if the model can do a reasonable prediction just outside the
data set boundaries. Obviously the different methods and pruning out of
variables will give different values of the coefficients. The inherent
problem is always what is the "best" out of many possibilities?


DAHeiser