Methods of validation of model with multivariate outputs

Three metrics for integrated validating models with multivariate outputs were developed based on classical hypothesis test and multivariate statistics method. These metrics could use the implicit correlation information between outputs for integrated validation, so false credibility evaluation because of neglected correlation information could be avoided. For commonly used statistical significance tests were based on normality assumptions, Box-Cox etc. Transformation methods were investigated in the case of non-normal data and extend the application. At last, numerical cases were offered to demonstrate that integrated validation could get a more accurate result than univariate validation. The application on 3 DOF kinematic model of one vehicle proves the necessity and applicative of methods.