@junpenglao thanks for that suggestion! I’ve tried this and several other combinations of specifying hyper parameters for the 3 variables and cannot get it to recover the parameters correctly. I’ve also tried not setting prior distributions for the sd parameters and just hard coding them in the a and b parameters with marginally better luck.
E.g.
1)
2)
I’m really stumped here! Shouldn’t this be a fairly easy model to sample?