Unreinforced masonry (URM) buildings constitute a large proportion of the existing building stock worldwide and are particularly vulnerable to earthquake damage. Their assessment and retrofitting, therefore, represent a critical challenge for earthquake engineering and risk reduction. Among the available modeling strategies, the equivalent frame model (EFM) has gained widespread acceptance in practice due to its balance between computational efficiency and ability to capture the global and local seismic response. However, the assumptions and simplifications underlying EFMs inevitably introduce uncertainties that remain insufficiently quantified, thereby limiting their reliability for informed decision-making. By adopting an EFM approach with the use of a recently developed three-dimensional macro-element formulation for the seismic assessment of URM buildings, accounting for the potential interaction of in-plane and out-of-plane responses, this thesis addresses this gap by systematically quantifying model uncertainties and their impact on predictions. For achieving this, the first database of full URM buildings tested on shake-tables was assembled, providing ground-truth evidence of global and local seismic behavior. Therefore, the experimental results were compared against nonlinear dynamic analyses of EFM-based models, enabling a direct quantification of modeling errors. Uncertainty was characterized at multiple levels: (i) the prediction of global building response, and (ii) the prediction of local response in piers, spandrels, and gables. The findings demonstrate that while the adopted EFM approach provides a useful tool for capturing global seismic performance trends, notable uncertainties arise in the prediction of local damage patterns. The magnitude and variability of these uncertainties were quantified, offering engineers a transparent understanding of the reliability of EFMs in practical assessments. In addition, the thesis evaluates the capability of EFMs, under commonly adopted modeling assumptions, to predict the extent and distribution of element-level damage. This reliability assessment enhances the knowledge of where EFMs can be confidently applied, and where further refinements are required. Ultimately, this research delivers a well-founded quantification of model uncertainties for URM seismic assessment, and proposes recommendations for incorporating these uncertainties into engineering practice. By doing so, it strengthens the basis for performance-based decision-making and provides a roadmap for improving existing modeling strategies. Beyond practical contributions, the work also highlights areas where future research could enhance the predictive capacity of EFMs, thereby advancing the safe and resilient use of URM buildings in seismic regions.