I am running a model when the response variable (behaviour) is an ordinal factor (levels: $0,1,2,3,4$) and I am testing the relationship with two fixed effect factors (year: $A,B,C,D$; colony; $X_1$,$X_2$).
I want to extract the estimates of the effect of year and colony on behaviour. In a normal GLM, I understand that this is simple, but for an ordinal model, I am not sure how to incorporate the cut points for the ordinal levels with the fixed effect estimates. From reading about ordinal models, it seems that maybe you get a matrix of probabilities such that you get for colony a $2 \times 5$ matrix and for year a $4 \times 5$. Am I understanding this correctly? And if so, it does not seem possible to extract any kind of residual value for behaviour controlling for fixed effects?