2.6. Summary

Both boosting and partial least squares for linear discriminant analysis can be easily implemented using widely available SAS modules. The theoretical properties and performance of each of these methods have been thoroughly researched by a number of authors. In this chapter, we have provided additional details of each method, have provided code to implement each method, and have illustrated their application on a typical drug discovery data set. Undoubtedly, as data structures become more complex, methods like thesebecome extremely valuable tools for uncovering relationships between descriptors and response classification.