The issues of random and systematic errors in thermodynamic models are examined. The general objective of this dissertation is the development of methods to identify and quantify the role of random and systematic errors in thermodynamic models and how they affect process design and simulation. Methodologies based on Monte Carlo approaches are shown to be good tools for sensitivity and uncertainty analyses of thermodynamic models. They can be successfully used to study the effect of random and systematic errors in a wide variety of situations. In general, more information about the regression and the model is generated through the use of the proposed methods. Specifically, the sensitivity of the binary interaction parameters for the NRTL model to the experimental data source and the effect of the data type (i.e. binary vapor-liquid and liquid-liquid data or ternary liquid-liquid data) on the regression of the UNIQUAC parameters were studied. The results show that these effects are very significant on the accuracy and precision of the models when applied to process performance evaluation and simulation. Traditionally, model evaluations are based only on the statistical estimators generated from the regression procedures, which do not provide a reliable evaluation most of the time for nonlinear models. Other issues such as the method used for parameter space estimation for sampling purposes and the regression approach used also significantly affect model evaluations, and they are explored in this project. An approach called Equal Probability Sampling (EPS) was developed for sampling the parameter space of nonlinear regression models with correlation among the parameters. The proposed EPS technique shows a much more realistic or practical interpretation of the uncertainty present in the regression of thermodynamic models. The parameter correlation is incorporated directly in EPS rather than using approximated pairing procedures. With respect to the regression approach used to obtain the binary parameters of thermodynamic models, a new approach called Inside-Variance Estimation Method (IVEM) is presented. This method is based on the maximum likelihood principle and involves the recomputation of the variance for each iteration of the optimization procedure. The IVEM procedure produces better results than traditional least squares regression procedures, including maximum likelihood formulations with a priori defined variance-covariance matrix. The improvement in the regression results obtained using the new approach for the objective function results in better quantification of the thermodynamic modeling errors, making models like UNIQUAC more reliable for prediction and simulation. A general approach based on Monte Carlo simulation was developed to analyze the effect of systematic and random errors on computer models. The method was applied to thermodynamic models to study systematic and random error propagation on chemical process design and simulation. The results from the case studies presented show that the approach is able to distinguish which error type has stochastic dominance on the uncertainty propagation through the computer model. It was concluded that systematic errors can play a significant role in error propagation and therefore in uncertainty analysis. Additionally, it was observed that the effect of random and systematic errors on the uncertainty propagation is not additive, meaning that the error type with stochastic dominance defines the uncertainty propagated to the predicted process performance or output variables of the model. Finally, the general objective defined for this project was attained to a very significant extent and suggestions are given for further research in this area.