How to set a prior on a collection of effects in a brms model

Modeling
1

I’m looking to make a model such as the following, but with a possibly unusual prior:

Outcome ~ Intercept + A + B + C + D + E

You could consider A through E as being like a tag or feature that the item has. However, I think there is good reason to believe that knowing the effect of one feature, or the typical impact of having some such feature, can inform the expectation of the effects of another feature. Hence, I am looking to make a hyperparameter that captures something like the average or typical effect of having a feature, and then additional feature-specific deviations. Ultiately this could also enable the postulation of a new feature ‘F’, which we could draw from the overarching distribution of likely feature effects.

How can such a model be achieved in brms, if it can be?

2

As an additional note, I think it’s possible that just logically, the average effect is going to be 0, so it may be that the prior is more on the SD of the specific effects or something like that?