Analyzing the interactional term in brms

Modeling
1

Hi everyone. I just learned how to do the Bayesian ordinal logistic regression model after reading Bürkner & Vuorre’s (2019) tutorial. I followed their steps, tried different models, compared them, and found that the following model with the interactional terms has the best fit:

model<-brm(formula=Scale~Condition*Language+(1|Subject)+(1|Item),data=data,family=cumulative(“probit”))

Scale is the Likert scale from 1 to 9
Condition is the categorical variable with four levels: Condition_a (used as the reference level), Condition_b, Condition_c, Condition_d
Language is the categorical variable with two levels: Language_English (used as the reference level) and Language_Spanish

Honestly, I don’t know how to further analyze the interactional term in Bayesian modelling. I mean, if I would like to treat Condition as the moderator, and see at each level of Condition whether Language has an effect, what should I do here?

I saw a similar post on this forum where someone suggested using hypothesis() function. Honestly, after reading that post and the usage of hypothesis(), I am not quite sure that I fully understand it.

For instance, if I use

hypothesis(model, (“Condition_b + Condition_b*Language_Spanish = 0”))

The result looks like this:

Hypothesis Tests for class b:
Estimate Est.Error CI.Lower CI.Upper Evid.Ratio Post.Prob Star
3.88 0.21 3.48 4.3 NA NA *

What is being tested here? How do I interpret these numbers?

2

Howdy. You can start by using conditional_effects() to view plots of the outcome at various levels of the predictors. See conditional_effects() in the brms manual for examples of using it to view interactions.
You can do the same thing yourself by creating datasets with the predictors at various values and then using fitted() to view the mean values of the outcome at each combination of predictors.
I’ve never used hypothesis(), so I can’t help you there.