Recalibrating single-study effect sizes using hierarchical Bayesian models