When we have developed a multivariate calibration it is essential that the performance of the calibration is tested with a set of samples which are typical and independent. This set has several names but I prefer "Validation set". If you are satisfied with the predictions on these samples then you will daim that you have "Validated" the calibration, which suggests that you expect it to continue to give useful results in the future.
For full article Click Here
Independence rules (or Rules for independence) : Validation and Calibration