A toolkit for cross-validation: The R package cvTools

The idea of cross-validation is simple and easy to implement: split the data into several blocks, leave out one block for model estimation, and predict the values of the left-out block. Those predictions are then used to compute a certain prediction loss function. Even though the basic procedure is simple, some additional programming effort is necessary for more complex procedures such as repeated (double) cross-validation, or using cross-validation to select the optimal combination of tuning parameters. While many packages for computing regression models already offer functionality for cross-validation, different packages use different interfaces and the returned objects have a different structure. Furthermore, developers often copy and paste the basic code skeleton when implementing cross-validation for different models, which complicates maintaining the code.