The high expense of vehicle testing leads to the desire to use analysis as much as possible to get an understanding of rollover resistance. There is no shortage of computer programs which claim to be capable of simulating rollover events. The challenge is in providing reasonable parameters for the simulation and in understanding the range of validity of the computed results. This paper presents various mathematical models which bear on rollover, starting with a very simple quasi-static model which requires knowledge of only a few measured vehicle parameters. It then deals with simulations of increasing complexity and increasing need for measurements to provide accurate and detailed input data. The paper closes with remarks on the relationship between simulation and vehicle testing for rollover resistance.