Analytical models are commonly used for the prediction of fiber-reinforced polymer (FRP)-to-concrete bond strength. In order to improve the precision and reliability of such models, extensive databases from different experimental research are used. Such databases are referred to as clustered databases, where data from individual experiments are organized into groups. Due to correlation effects between data within a group, also known as intra-cluster correlation, classical regression methods are inefficient in capturing data trends and utilization of appropriate techniques are mandatory. The mixed-effects regression method is an efficient tool for analyzing clustered databases. In this study, the effect of considering the hierarchal structure of data in calibrating FRP-to-concrete bond strength is studied by calibrating a simple model form and comparison with commonly used models. It is elaborated that considering data correlation by utilizing the mixed-effects regression method, significantly decreases a model’s uncertainty and improves its reliability.
