This paper introduces a freeware program for models analysis, which implements a robust method for parameter estimation of nonlinear models, able to detect outliers. A new class of criteria is proposed in order to achieve an optimal experimental design. Finally a strategy for model discrimination is presented. Since this is a critical point. the correct meaning of statistical tests and regions of confidence is proposed. At last, it is explained why it is better to avoid any approach that makes use of probability to discriminate among models. (C) 2008 Elsevier Ltd. All rights reserved.