This PR proposes some changes to inferencedata1. The main ones are:
Instead of _symbol, use the suffix _var for keyword args. This is consistent with the from_cmdstan, from_pystan, and from_cmdstanpy methods in Python ArviZ and would be more familiar for a Python Stan user coming to Julia.
The _var keyword args can now take Symbol, or an iterator of Symbols.
Add a predictions_var keyword arg
Add split_nt and split_nt_all convenience methods and refactor to use them for splitting the samples named tuple into the various group named tuples
Support other numbers of warmup or sample draws than 1000.
This PR proposes some changes to
inferencedata1
. The main ones are:_symbol
, use the suffix_var
for keyword args. This is consistent with thefrom_cmdstan
,from_pystan
, andfrom_cmdstanpy
methods in Python ArviZ and would be more familiar for a Python Stan user coming to Julia._var
keyword args can now takeSymbol
, or an iterator ofSymbol
s.predictions_var
keyword argsplit_nt
andsplit_nt_all
convenience methods and refactor to use them for splitting the samples named tuple into the various group named tuplesHere's an example output, using the notebook at https://github.com/StanJulia/Stan.jl/blob/master/Examples_Notebooks/InferenceObjects.jl
Note that in the next breaking release of InferenceObjects, the dimension orders of arrays will change (https://github.com/arviz-devs/InferenceObjects.jl/pull/40), and the default indices for all dimensions will be the
axes
of the underlying arrays (https://github.com/arviz-devs/InferenceObjects.jl/pull/39; so after splitting samples from warmup, no reindexing will be needed)