Releases: arviz-devs/arviz_example_data
v0.3.0
Update eight schools data (#11)
The non-centered eight schools observations and predictions now use the school dimension to match centered eight schools. From both centered and non-centered eight schools, the old scores variable in constant_data was removed (duplicate of obs in observed_data), and the variable sigma was added to constant_data.
Full Changelog: v0.2.0...v0.3.0
v0.2.0
Update the existing rugby example data
The rugby example data is updated to be up to date with ArviZ completeness (not really best practices).
The model stays the same, the variables stay the same, and the groups that were already there also
stay mostly the same, so nothing should break. The only change to existing groups is adding extra indices,
a new xarray feature that can help with groupby-like behaviour in plot_ppc and the like.
The main change therefore is the addition of the log_prior and unconstrained_posterior groups, which were
recently added to the schema and will help generate examples of the functions that do use them.
Currently things are a bit cumbersome in such cases, for example psens function manually computes the log prior using xarray-einstats.
Add a new rugby_field example data
This adds a new example data that will mostly be used for examples and maybe testing on arviz-plots and the other prototypes. The highest dimensional variable we have in all existing datasets is chain, draw, extra_dim, which is a bit limited in terms of facetting/aesthetic mapping combinations.
This new dataset extends the rugby one (which I believe will make it easier for users to understand and also allows small dimension size team=6 and field=2 while being realistic and intuitive) to have a couple of 4d variables chain, draw, team, field and also to have 3d variables with both chain, draw, team and chain, draw, field.
v0.1.1
What's Changed
- Update 8-schools models and InferenceData by @michaelosthege in #8
- some fixes to updated 8 schools models by @OriolAbril in #9
Full Changelog: v0.1.0...v0.1.1
