Validity of model simplification

The validity of model simplification is defined in statistical terms. Two models, the reference model and the simplified model, are compared. The simplified model is called 1- gamma acceptable if it is not rejected with probability 1- gamma , by a statistical test which uses data generated by the reference model. The model is called a valid simplification on the validity level nu if the ratio of the probability of its acceptance to the probability of acceptance of the reference model is equal to nu . This definition takes into account a possibility of observing the reference model rather than the data. The application of the Kolmogorov test to validity testing is discussed.<
>