Bump Stan on Julia 0.6 to v2.0.3

Added algorithm=fixed_param to Sample().

Tag Stan.jl v3.0.1 (Julia 0.7-)

Added DiffEqBayes based examples.

Tag Stan.jl v2.0.2 (Julia 0.6)

Added DiffEqBayes based examples. Dropped support/testing on Julia 0.5.

Tag Stan.jl 3.0.0

Requires Julia 0.7-

Tag Stan.jl v2.0.1

Fixes an issue with thinning and clarifies arguments to stan() docs. Thanks to the contributors!

Next major release, currently in branch 3.0.0 is Julia 0.7- only.

Stan.jl v2.0.0 Compatible with Julia 0.5 & 0.6-RC1

Stan-jl-2.0.0 (Stan.jl v"2.0.0") which is compatible with both Julia 0.5 and Julia 0.6.

It does contain 4 breaking changes compared to Stan.jl v"1.1.0" and is thus a :major level upgrade:

1. The addition of the useMamba keyword argument to Stanmodel().
2. The update keyword argument to Stanmodel() has been renamed num_samples.
3. Passing in parameter initial values (used to be in Stanmodel(), now in stan() using identical format to the data argument.
4. stan() returns a tuple (return code, sim).

The sim object is either a Mamba.Chain or an Array depending on the value of the useMamba setting in Stanmodel(). By default useMamba is set to true. If conversion to Mamba chains is not desired or possible (e.g. initially for testing new Julia versions when Mamba and Gadfly might not be available yet), set useMamba to false in the call to Stanmodel().

Consolidate changes to Stan.jl master for Julia 0.5

This release (Stan.jl v"1.1.0") contains all updates to Stan.jl master until May 1st 2017. It is not yet compatible with Julia 0.6. Stan.jl v"1.1.0" does support initialization of parameters in the Stan program parameter block through an init arguments in Stanmodel(). This will change in Stan.jl v"2.0.0" where this argument will move to stan() and the input format - a dictionary - is identical to the data argument (for observed data).

Almost simultaneously I will release Stan-jl-2.0.0 (Stan.jl v"2.0.0") which is ready for Julia 0.6. It does contain 2 breaking changes compared to Stan.jl v"1.1.0" and is thus a :major level upgrade. The first breaking change is described above. The second is that stan() will return a tuple (return code, sim).

Update Julia lower bound to 0.5.0-rc3 to match Mamba

Updated README accordingly and .travis.yml to only test on Julia 0.5

Stan.jl interface for Julia 0.5

Version 1.0.0

1. Tested with Stan 2.11.0 (fixed a change in the diagnose .csv file format)
2. Updated how CMDSTAN_HOME is retrieved from ENV (see also REQUIREMENTS section below)
3. Requires Julia 0.5
4. Requires Mamba 0.10.0
5. Mamba needs updates to Graphs.jl (will produce warnings otherwise)

Stan.jl interface for Julia 0.4

Version 0.3.2

1. Cleaned up message in Pkg.test("Stan")
2. Added experimental use of Mamba.contour() to bernoulli.jl (this requires Mamba 0.7.1+)
3. Introduce the use of Homebrew to install CmdStan on OSX
4. Pkg.test("Stan") will give an error on variational bayes
5. Last version update for Julia 0.4