This paper presents a study on dimensional variations and tolerance analysis and synthesis for polymer matrix fiber-reinforced composite components and assemblies. A composite component dimensional variation model was developed with process simulation based on thermal stress analysis and finite element analysis (FEA). Using the FEA-based dimensional variation model, the deformations of typical composite structures were studied and the regression-based dimensional variation models were developed. The regression-based dimensional variation models can significantly reduce computation time and provide a quick design guide for composite products with reduced dimensional variations. By introducing a material modification coefficient, the comprehensive regression models can handle various fiber and resin types and stacking sequences, which eliminates the complicated, time-consuming finite element meshing and material parameter defining process. A structural tree method (STM) was developed for rapid computation of composite assembly dimensional variations resulting from deformations on individual components, as well as the deformation of composite components with complex shapes. With the STM and the regression-based dimensional variation models, rapid design optimization was conducted to reduce the dimensional variations of composite assemblies. Cost-tolerance functions were developed using a fuzzy multiattribute utility theory based cost-estimation method. Based on the developed dimensional variation and cost-tolerance models, composite assembly tolerance analysis and synthesis were performed in this study. The exploring research work presented in this paper provides a foundation for developing practical and proactive dimensional control techniques for composite products.