Brm_multiple but in the same model, multiple features

I would like to express a model with multiple feature tested, and herarchical relationships and contrains across parameters.

y1 ~ 1
y2 ~ 1
y3 ~ 1
sigma_1,2,3 ~ Intercept_1,2,3 * b   # or more specifically  -> log(sigma_1,2,3) ~ inv_softmax(Intercept_1,2,3) * b
[Intercept_1,2,3] = simplex

Is this possible in brms?

It’s possible but as far as I know it’ll be quite verbose and not too pretty. If you really want to, you can use the nonlinear syntax to define the quantities that are reused in predicting multiple parameters and the relationships according to which these predictions should be made.