Data Science: Similarity Least Squares (SLS^TM) + physics + Statistical Design of Experiment (DOE)

            A New Paradigm for Analysis and Management of Complexity

Model-instability syndrome (ref clinical study):

DOE principles identify a set of critical profiles that assure model stability.
When (super) critical profiles are split between a validation data set and a training data set, error rates of the consequential unstable model are inflated.
 
This is a common problem in current clinical modeling because such critical points are not identified. Consequently, the training set must be larger than the validation set to inherently reduce the expected number of critical points (hidden unknowns) that will end up in the validation set and hence, reduce the extent of distortion.

Conversely, with SLS^TM modeling the training set is inherently much smaller than the validation set because critical profiles are identified by DOE concepts.

Copyright of James M Minor, July 4, 2004.
For problems or questions regarding this web contact email jmin007@yahoo.com
Last updated: June 20, 2013.