A “Leaps and Bounds” Algorithm for Variable Selection in Two-Group Discriminant Analysis
An algorithm that identifies the variable subsets with most discriminatory power (in a predictive sense) is proposed. This algorithm minimizes parametric estimates of the error rate among all the possible variable subsets, evaluating only a fraction of the total number of subsets. The computational feasibility is illustrated by simulation experiments.
KeywordsDiscriminant Analysis Variable Selection Techniques Optimization Techniques
Unable to display preview. Download preview PDF.
- Huberty, C.J. 1994. Applied Discriminant Analysis, New York, NY: Wiley.Google Scholar