Is there a subtle bug in the splines in stan case study?
In section “Location of Knots and the Choice of Priors” it states to use a N(0, tau) prior on a, but in the code it looks to be N(0, 1). Full code is on github here.
The linked model does indeed imply a N(0, 1) prior on a and then a N(0, tau) distribution on the subsequent innovations. It looks like Milad forgot the * tau needed in the non-centered parameterization of a.
Milad corrected the case study and I updated the case study on the web site: