bggum: Bayesian Estimation of the Generalized Graded Unfolding Model

Standard IRT models assume that higher trait levels always produce higher endorsement probabilities — an assumption that fails for “ends against the middle” items, where respondents with opposing extreme preferences respond identically. bggum implements the generalized graded unfolding model (GGUM) via Metropolis-coupled Markov chain Monte Carlo, recovering ideal points even when opposites respond alike. Includes post-processing, parameter estimation, and plotting utilities. Companion to Duck-Mayr and Montgomery (2023), Political Analysis (Warren Miller Prize, best paper in Political Analysis).