A. Lorenzo Arribas, M. J. Brewer, A. M. Overstall
Model assessment in ordinal response models entails many practical issues.Traditional goodness-of-fit measures available for linear fixed and random-effects models and diagnostic plots of individual residuals are generally unavailable and difficult to implement for these models.Main problems are associated with the discrete nature of the data.Residual diagnostics for models with ordinal outcomes are generally accepted as not well developed.Challenges include: they do not provide only one value per subject, they do not reflect the overall direction of the observed values,they are not monotonic with respect to the observed values for those with the same covariates,they do not have a range of possible symmetric values with respect to zero and they do not have expectation zero.We propose the use of randomised quantile residuals as a solution to the challenges derived from existing approaches and we build a framework for assessment of goodness of fit in the ordinal response models context.
Palabras clave: ordinal response data, goodness of fit, residualsProgramado
L08.3 Sesión de la Red Nacional de Bioestadística
5 de septiembre de 2016 15:40