Simulation of Automotive Warranty Data

This thesis will investigate the prediction of the number of claims in a two dimensional automotive warranty claim model for the case of minimal repair.The method involved fitting marginal distributions for age of claim and mileage of claim seperately. Next, various copulas were fitted to establish the correlation between age and mileage, and assessed for fit. The Gumbel copula is chosen as optimal. From this Gumbel copula, a simulation of warranty claims is undertaken. The method produced a good fit for claim age but performed less well for claim mileage, due to the asymmetry of the correlation between mileage and age. Further research directions to improve the accuracy and usefulness of this model are suggested.