Convert stats and diagnostics to new API #232
Conversation
ColCarroll
force-pushed
the
xdiagnostics
branch
from
c1293c7
to
59038f9
Compare
ColCarroll
changed the title
|
I think this is now ready to go. There's some missing documentation, but also this will allow us to delete most of
|
arviz/stats/stats.py
Outdated
| raise ValueError('The number of observations should be the same ' | ||
| 'across all models') | ||
| raise ValueError( | ||
| "The number of observations should be the same " "across all models" |
There was a problem hiding this comment.
Being picky, redundant parentheses here?
There was a problem hiding this comment.
Fixed - I actually ran black on the file as an experiment - this might have been part of the result.
ColCarroll
force-pushed
the
xdiagnostics
branch
from
59038f9
to
0f6ccd4
Compare
|
LGTM |
|
|
||
|
|
||
| __all__ = ['effective_n', 'gelman_rubin', 'geweke'] | ||
|
|
||
|
|
||
| def effective_n(trace, varnames=None, round_to=2): | ||
| def effective_n(data, *, var_names=None): |
There was a problem hiding this comment.
I didnt know you can use * as input - what does it do?
There was a problem hiding this comment.
It prevents people from using positional arguments, in this case in excess of data
|
LGTM as well |
These changes get the test suite for
diagnostics.pyto pass, with everything being xarray based. Now will need to work onstats.py, which will remove (most of) the pandas requirements. I am working on that, but will probably need help for some of the functions (#231) .Note that most functions in
diagnostics.pywork with either a dataset, or a numpy array.