Bounded responses, either continuous or discrete, are common in many fields, such as ecology, economics, and biomedical sciences. Standard regression models often fail to capture features like bimodality, heavy tails, overdispersion, or excess zeros, which frequently arise in these contexts. The FlexReg package, which provides a unified Bayesian framework for regression modeling of bounded outcomes, addresses all these challenges through flexible distributional assumptions and robust estimation based on Hamiltonian Monte Carlo implemented via the Stan language. The package implements beta-type and binomial-type models, along with their flexible and variance-inflated extensions, and allows for handling values on the boundary of the response support, when needed. Beyond model fitting, FlexReg includes tools for convergence diagnostics, posterior summaries, residual analysis, and predictive checking. By integrating and enriching recent methodological advances within a user-friendly Bayesian framework, this work delivers a computational infrastructure that facilitates the application of flexible regression methods for bounded data.
