The caret package in R, short for classification and regression training, contains several impressive tools for the study of prediction and classification methods. For a given prediction problem there may be several competing approaches. Arguably, one cannot know which approach is optimal until a thoughtful and sound empirical investigation is carried out on competing methods. The design of the caret package provides tools that unify the processes of data splitting, feature importance, model tuning, visualization, and performance comparisons. In this presentation we give a case study of how the caret package can be employed as a teaching tool to introduce multiple competing prediction and classification approaches.