Traditional validation of a response model is based on a 'hold-out' sample, which consists of individuals who are not part of the sample used in building the model itself. The validation results are probably biased and definitely incomplete - unless a resampling method is used. This article points to the weaknesses of the traditional validation, and then presents a new approach for validating response models using the bootstrap.


Related Articles:
1. Creating A Bootstrap Sample
2. When Data Are Too Large to Handle in the Memory of Your Computer 
3. A Simple Bootstrap Variable Selection Method for Building Database Marketing Models