Stan 2.3 released (RStan, PyStan, CmdStan)

[Bob Carpenter]

We're happy to announce RStan, PyStan and CmdStan 2.3.
Instructions on how to install at:

http://mc-stan.org/

Release notes below.

As always, let us know if you're having problems or have comments
or suggestions.

- Stan Development Team


v.2.3.0 (18 June 2014)
======================================================================

We had a record number of user-submitted patches this go around.
Thanks to everyone!

New Features
------------
* user-defined function definitions added to Stan language
* cholesky_corr data type for Cholesky factors of correlation matrices
* a^b syntax for pow(a,b) (thanks to Mitzi Morris)
* reshaping functions: to_matrix(), to_vector(), to_row_vector(),
to_array_1d(), to_array_2d()
* matrix operations: quad_form_sym() (x' *Sigma * x), QR decompositions
qr_Q(), qr_R()
* densities: Gaussian processes multi_gp_log(), multi_gp(),
and alternative negative binomial parameterization neg_binomial_2()
* random number generator: multi_normal_cholesky_rng()
* sorting: sort_indices_*() for returning indexes in sorted order by
value
* added JSON parser to C++ (not exposed through interfaces yet; thanks
to Mitzi Morris)
* many fixes to I/O for data and inits to check consistency and
report errors
* removed some uses of std::cout where they don't belong
* updated parser for C++11 compatibility (thanks to Rob Goedman)

New Developer
--------------
* added Marco Inacio as core developer

Optimizations
-------------
* turned off Eigen asserts
* efficiency improvements to posterior analysis print

Documentation
-------------
* Clarified licensing policy for individual code contributions
* Huge numbers of fixes to the documentation, including many
user-contributed patches (thanks!), fixes to parallelization in
CmdStan, Windows install instructions, boundaries for Dirichlet and
Beta, removed suggestion to use floor and ceiling as indices,
vectorized many models, clarified that && doesn't short circuit,
clarified von Mises normalization, updated censoring doc (thanks
to Alexey Stukalov), negative binomial doc enhanced, new references,
new discussion of hierarchical models referencing Betancourt and
Girolami paper,
* Avraham Aadler was particularly useful in pointing out and fixing
documentation errors

Bug Fixes
------------
* fixed bug in lkj density
* fixed bug in Jacobian for corr_matrix data type
* fix cholesky_cov_matrix test code to allow use as parameter
* fixed poisson_rng, neg_binomial_rng
* allow binary operations (e.g., < and >) within range constraints
* support MS Visual Studio 2008
* fixed memory leaks in categorical sampling statement, categorical_log
function, and softmax functions
* removed many compiler warnings
* numerous bug fixes to arithmetic test code conditions and messages,
including calls from
* fixed model crashes when no parameter specified
* fixed template name conflicts for some compiler bugs (thanks Kevin
S. Van Horn)

Code Reorganizations & Updates
------------------------------
* CmdStan is now in its own repository on GitHub: stan-dev/cmdstan
* consolidate and simplify error handling across modules
* pulled functionality from CmdStan command class and PyStan and RStan
into Stan C++
* generalized some interfaces to allow std::vector as well as Eigen
for compatibility
* generalize some I/O CSV writer capabilities
* optimization output text cleaned up
* integer overflow during I/O now raises informative error messages
* const correctness for chains (thanks Kevin S. Van Horn)

[Marco Inacio]

How about including CmdStan in source-repos.html file?

[Bob Carpenter]

I'm not even sure what the fix is in the lkj_density.
Anyone know what that was?

For any devs listening [stan-dev cc-ed] --- we need better commit messages
because it's what we use to decide how to do the release notes. Put
the issue number in the fix for for it and I will reference that.

There was a missing term in the correlation matrix Jacobian, though,
which could lead to a bit of bias. We weren't getting exactly uniform
distributions on correlation matrices. I doubt it'll matter for
what you're doing or we'd have made a much bigger deal out of it.

- Bob

On Jun 26, 2014, at 9:34 PM, Sebastian Weber <sdw....@gmail.com> wrote:

>
> My congratulations. I am really looking forward to start using these user-functions which will make my Stan code a lot better recyclable... and even force me write "nicer" Stan models...
>
> ... one note: If you write that a bug was fixed in lkj_density (or something with similar wide usage) I would appreciate if some note to its relevance could be given, i.e. is it minor or do I have to throw away my old inferences - just to be a bit extreme.
>
> So maybe a link to the ticket or git commit or better a comment about the impact of the bug would be very user-friendly here. I know that Stan is already one of the most user-friendliest softwares out there, but as Stan is used for more and more "serious" stuff, a bug here and there could matter to some of us.
>
> Many thanks for making the release again - looking forward to play with it!
>
> Cheers,
> Sebastian

> --
> You received this message because you are subscribed to the Google Groups "Stan users mailing list" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to stan-users+...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Ben Goodrich]

On Thursday, June 26, 2014 5:48:46 PM UTC-4, Bob Carpenter wrote:

I'm not even sure what the fix is in the lkj_density.
Anyone know what that was?

I think it referred to making lkj_corr_cholesky have the expected behavior instead of the behavior it has always had. But now it seems as if that code didn't get merged. I pushed it onto

https://github.com/stan-dev/stan/tree/feature/issue-413-cholesky-corr-type

Ben

[Jeffrey Arnold]

Just a heads up, the links on http://mc-stan.org/ under the Interfaces section still say 2.2.0.

On Thursday, June 26, 2014 5:48:46 PM UTC-4, Bob Carpenter wrote:

[Bob Carpenter]

I think you may just need to reload in the browser.
(The need to explicitly reload in Safari drives me
crazy.)

- Bob

> You received this message because you are subscribed to the Google Groups "stan development mailing list" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

[Jeffrey Arnold]

You're correct. Sorry about that.

[Kevin Van Horn]

The new manual says,

"The CmdStan documentation used to be part of this document, but is now its own standalone document. The CmdStan home page, with links to download and installation instructions and the manual is code and manual is http://mc-stan.org/cmdstan.html "

But going to that URL, there is no CmdStan manual. Downloading CmdStan, the only manual found is the Modeling Language Manual. The README.txt file says that

"Stan's user's guide and reference manual, including installation 

instructions and a quick start guide, are in:

  $STAN/doc/stan-reference.pdf"

But that is just the modeling language manual, and it contains nothing specific to CmdStan (like all the info it used to have on command-line parameters, how to build Stan, what tools you should have installed, etc.)

So it appears that this release is lacking any documentation for CmdStan.