Analysis of sensory ratings data with cumulative link models

摘要: Examples of categorical rating scales include discrete preference, liking and hedonic scales. Data obtained on these are often analyzed with normal linear regression methods or omnibus Pearson chi2 tests. In this paper we propose to use cumulative link models that allow for similar while respecting the nature observations. We describe how related tests they can lead more powerful in non-replicated setting. For replicated ratings data present a quasi-likelihood approach mixed effects both being extensions models. contrast population-average subject-specific interpretations based discuss different approaches settings, naive ignore replications expected be too liberal because over-dispersion.We depends whether experimental design is fully randomized blocked. latter situation stronger than over-dispersion adjusting approaches, provide even will given throughout methodology implemented authors’ free R-package ordinal.