Interval estimation for individual categories in cumulative logit models
Abstract
We derive and compute confidence intervals for probabilities of categories for ordinal responses in cumulative logit link models using frequentist, Bayesian, and bootstrap approaches. The three approaches are evaluated by their coverage probabilities and are illustrated for longitudinal assessment of graded heart valve prosthesis regurgitation following aortic valve replacement. Copyright © 2007 John Wiley & Sons, Ltd.