The background of this thesis is the GMMC Multi-objective Antenna Optimization project. In that project the rational RBF interpolation method was developed to create surrogate models suitable for antenna optimization. Proposed here is a new strategy for choosing evaluation points in antenna optimization problems, compatible with the rational RBF interpolation method. We introduce the concept of error indicators, which can be used to locate regions with potentially large differences between the objective function and the corresponding surrogate model. By selecting new evaluation points where the the error indicator attains its maximum we can adaptively improve the quality of the surrogate model. Starting with a set of different error indicators and antenna simulation data, we use statistical methods to compare their performance with respect to error reduction. Also, this selection procedure is designed so that it can easily be applied to other similar problems where one wants to choose between models that are influenced by chance.