ABSTRACT
In the case of a covariate with ordered categories, it is often expected that the outcome increases or decreases with the level of the category, that is, there is a trend in the relationship. This can be taken into account, for example, by coding the categories as 1, 2, 3, . . ., and then treating this covariate as a continuous one. However, this implies an a priori assumption about equal differences between neighbouring categories, that is, with the notation of the last section we assume
δ12 = δ23 = δ34 = . . . .
