What @Mauricio_Garnier-Villarre says sounds like the right thing.
I was worried about the sum to zero constraint but I think I was wrong to be here. You only need N - 1 parameters to define a length N vector with a sum to zero constraint, but, if what you have works, don’t worry about this and just roll with what you have. There was a thread where people talked quite a bit about these sum to zero things but I don’t think anything was absolutely conclusive: Test: Soft vs Hard sum-to-zero constrain + choosing the right prior for soft constrain.