Properties of the SCOOP Method of Selecting Gene Sets
Yushi Liu · OhioLink ETD Center (Ohio Library and Information Network) · 2010
We developed a new variable selection method, Shrunken Centroid Orthogonal Ordering Projection (SCOOP).This method was originally inspired by nearest shrunken centroid method by Tibshirani [76].Unlike the original method, which shrinks along the natural axis, SCOOP defines a new augmented discriminant space (ADS) containing the correlated discriminatory information from both the between epoch variance covariance matrix and the within epoch variance covariance matrix.In particular, SCOOP can produce an order of variables according to the novel discriminative information criterion that balances discriminatory and correlative information.In the application, microarray measurements of gene expression were made in the pituitary, liver and ovary of rainbow trout (Oncorhynchus mykiss) over the course of their one year spawning cycle, and new methods were developed I am indebted to my advisor Dr. Joe Verducci, an outstanding professor in data mining research.When I was a beginner in the statistical field, Joe guided me step by step, providing all the necessary and detailed answers to my questions.Without him, this thesis can not be done.He is a perfect example for me to follow in real life.I greatly appreciated our collaborators: