Journal of Data Science

Abstract: Supervised classifying of biological samples based on genetic information, (e.g., gene expression profiles) is an important problem in biostatistics. In order to find both accurate and interpretable classification rules variable selection is indispensable. This article explores how an assessment of the individual importance of variables (effect size estimation) can be used to perform variable selection. I review recent effect size estimation approaches in the context of linear discriminant analysis (LDA) and propose a new conceptually simple effect size estimation method which is at the same time computationally efficient. I then show how to use effect sizes to perform variable selection based on the misclassification rate, which is the data independent expectation of the prediction error. Simulation studies and real data analyses illustrate that the proposed effect size estimation and variable selection methods are com petitive. Particularly, they lead to both compact and interpretable feature sets. Program files to be used with the statistical software R implementing the variable selection approaches presented in this article are available from my homepage: http://b-klaus.de.