Ordinal scale analysis
Both the lesion healing scores and the appetite scores were recorded on
a four-point ordinal scale, not a quantitative scale, so appropriate
ordinal categorical methods were required for this analysis (Agresti,
2002). In addition, walking was also modelled on a three-point ordinal
scale: (1: immobile; 2: walking with difficulty; 3: walking normally).
An ordinal logistic mixed model was fitted to each data set with fixed
effects for Treatment, Day, Breed and Age (covariate), and a random
effect for the individual Animal ID. A Treatment × Day interaction was
included in each of the models, to allow for a different shaped time
course for each treatment. Note that the fitted model returned a set of
model-based probabilities of obtaining each possible score (1 through 4,
or 1 through 3), for the particular combination of terms in the model.
The model was fitted using the clmm function in the ordinal package of R
(Christensen, 2019), and probability estimates obtained using the
emmeans (Lenth, 2020) and RVAideMemoire (Hervé, 2020) packages in R.