The safety analysis of nuclear power plants design and operation is moving towards a realistic approach in which the simulations performed using best estimate computer codes must be accompanied by an uncertainty analysis. Uncertainty analysis allows considering the uncertainties in plant conditions and physical models in the simulation of the transient, known as the Best Estimate Plus Uncertainties (BEPU) approach, which is accepted by the deterministic safety analysis regulatory authorities. Most of the methods of uncertainty analysis are based on the propagation of uncertainties from input parameters to output magnitudes of interest. These outputs are calculated using computational codes, and to characterize them correctly with their uncertainty, a large number of simulations is often required. When this is prohibitive, due to the high computational cost, one should use methods that obtain a good characterization of the output uncertainty with a limited number of simulations. Usually, in the nuclear field, non-parametric methods are used for estimating tolerance regions, such as Wald method. The present work shows alternative approaches to Wald method, parametric and non-parametric, in constructing regions of tolerance associated with safety magnitudes of interest (e.g. the peak cladding temperature and maximum local oxidation in a Loss of Coolant Accident analysis). A case of application of large break loss of coolant accident in the cold leg of a pressurized water reactor is presented.