Discrimination Under Covariance Structure: A Consideration in DNA Microarray Technology

• Activity Number: 15
• Type: Contributed
• Date/Time: Sunday, August 11, 2002 : 2:00 PM to 3:50 PM
• Sponsor: ENAR
• Author(s): Anuradha Roy*+ and Ravindra Khattree
• Affiliation(s): Oakland University and Oakland University
• Keywords: classification ; covariance structure ; microarray

Abstract

DNA microarray technology is an effective technology to analyze experimental differences in populations of biomolecules by monitoring the levels for thousands of genes simultaneously. For example, for classification of tumors using gene expression data, the data set generally has a very large number of variables (genes) 5,000- 10,000 relative to the number of observations (tumor samples), say 100. The conventional classification rules fail as the number of variables are much more than the number of unit. To reduce the dimensionality of the estimation procedure we consider an approach by introducing some kind of covariance structure on the variance-covariance matrix of the observations on each unit. The computational difficulties and possible ways out are also suggested.