Marginal_effects for multivariate models with mixed continuous/ordinal responses

#1

I have a brms model with 8 continuous and 3 ordinal responses, and I would like to make a marginal_effects plot for one binary predictor per response. When I run the code, I get a warning that recommends using categorical=TRUE for the ordinal responses. However, doing this throws an error, since I have continuous responses. Is there a workaround here?

Alternatively, is there some way to only calculate the effects for a subset of responses?

#2

You can savely ignore the warning. It is just issued because brms recognizes that some of your response variables is ordinal.

#3

Thanks for the quick reply.
Is there some way to getting the probability of belonging to each ordinal level with this? I think this would be more interpretable.

#4

Ah I see what you mean. You should use the resp argument to plot results only for the selected responses.

#5

Ah, I must not have seen that in the documentation. Thanks!