Recently I was plotting the conditional effects of the following model:
fit ← (cat1, cat2, cat3) ~ Var1 + Var2 + Var3 + Var1 * Var2
In order to plot the interaction, I used the following code:
conditions <- brms::make_conditions(fit, vars = "Var1") plot(brms::conditional_effects(fit, ask = FALSE, categorical = T, conditions = conditions))
This provides the graph I was looking for but also provides a graph for the interaction between
Var3, despite this interaction term not being specified in my model.
When looking at the documentation for
conditional_effects it states:
effects manually, all two-way interactions may be plotted even if not originally modeled.
My question is, where do the values (estimates, CIs, etc.) come from if these interactions were not specified in the model? Are all interactions estimated even if they are not specified? Are the values in these graphs reliable?