For optimal design of a deep-sea ocean mining collector system, based on self-propelled mining vehicle, it is imperative to develop and validate the dynamic model of a tracked vehicle traveling on soft deep seabed. The purpose of this paper is to evaluate the fidelity of the dynamic simulation model by means of response surface methodology. Various statistical techniques related to response surface methodology, such as outlier analysis, detection of interaction effect, analysis of variance, inference of the significance of design variables, and global sensitivity analysis, are examined. To obtain a plausible response surface model, maximum entropy sampling is adopted. From statistical analysis and prediction for dynamic responses of the tracked vehicle, conclusions will be drawn about the accuracy of the dynamic model and the performance of the response surface model.