Abstarct: Developing models for predicting phenomena affected by human behaviour requires specific measures for considering the uncertainties affecting these phenomena. Travel demand is also one of these uncertain phenomena and one of the common shortcomings in this field studies is the failure of sensitivity analysis and uncertainty survey of these models. Therefore, to construct the travel demand estimation models, analysts have always used tools that can consider the uncertainties in these models. Generally, having a range of probable values in accompany with the probability of each occurring, is more important compared with having only one output (point estimate) with the highest probability of occurrence. This study aims to address two main objectives by performing a sensitivity analysis: 1) To quantitatively measure the sensitivity of travel demand models; 2) To rank the input variables which play the most role in model sensitivity. The Monte Carlo simulation method, which is a pseudo-random technique, is commonly used to investigate the uncertainty. But in this method, random numbers are less uniform and a higher number of iterations is needed. The Monte Carlo method, which is a pseudo-random method, is commonly used to investigate uncertainty. But in this method, random numbers have less uniformity and a higher number of repetitions (i.e. runs) is required. To overcome this shortcoming, the Latin Hypercube Sampling (LHS) and quasi-random methods of Halton and Sobol were investigated and their performance is compared with the Monte Carlo method. Furthermore, the normal, log-normal and triangular distributions and different amounts of the variation coefficients were used in the numerical simulation process. Investigation of the uncertainty of Mashhad travel creation models revealed that the Halton method despite the greater uniformity in the single space, produced the same uncertainty compared with other methods in the models and actually the different sampling methods do not have a significant effect on the uncertainty results. Moreover, the uncertainty of travel attraction models is greater than that of the travel production models. In some cases, the uncertainty of travel attraction models is greater than the assumed initial values. The sensitivity analysis indices calculated for the point and simulated data provide almost the same ranking of the sensitivity for the variables of the models of travel production and attraction and no numerical simulation is needed to rank sensitive variables.