t-tests, ANOVA, and linear regression all assume that the response is measured on an interval scale, so that the differences between adjacent values have the same meaning across the scale. This assumption is often violated in practice, which can lead to inaccurate conclusions. The proportional odds and proportional hazards models are ordinal regression models that are only sensitive to the order of the observations, not the specific values assigned to the categories. They are used to compare the distributions of JAR scores among products and are performed simultaneously. The proportional odds model (POM) is skew-symmetric, so that reversing the order of the scale simply changes the sign of the mean, while the proportional hazards model (PHM) is asymmetric, so that reversing the order of the scale changes both the order and the sign of the estimate.