======================
* Complete rewrite of the underlying model creation engine (kabuki) to
allow for more flexible model creation including transforms. This
enabled development of a new HDDM default model without explicit
parameter bounds.
* Group mean distributions are now Gibbs sampled and group variability
distributions are now slice sampled leading to much improved
convergence and mixing.
* MAP approximation of hierarchical models for better initialization.
* Improved documentation (check out the `How-to`_ section).
* Chi-square fitting using the Ratcliff quantile method.
* Posterior predictive checks.