Update some examples to use new arviz #822
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
OriolAbril
left a comment
OriolAbril
left a comment
There was a problem hiding this comment.
other than the extra files with .md instead of .myst.md extension it looks good
| all += perct_per_class | ||
| print(perct_per_class) | ||
| all | ||
| az.plot_ppc_pava(idata_t, data_type="categorical"); |
There was a problem hiding this comment.
My main comment is probably not actionable but might be worth to keep in mind for arviz development. I am not sure I can see this as it is a comparison between this figure and one a few sections higher. Does plot_ppc_pava support dict input to show both models, if not, do you think that is useful to add?
There was a problem hiding this comment.
I think the only ones supporting more than one model are plot_forest and plot_dist. Not sure how useful, it could be ok for plot_ppc_pava and maybe plot_ppc_pit. Maybe something more general is to have something similar to combine_plots but that works for different data. So we can create a single figure with one model per row or column, so comparisons are easier to see and present.
examples/bart/bart_introduction.md
Outdated
There was a problem hiding this comment.
this one also seems to be an extra file. There might be something broken in the jupytext settings :/
| ) | ||
| ax[1].set(title="Posterior Predictive Check (test)", xlim=(0, 1_000)); | ||
| ```{code-cell} ipython3 | ||
| az.plot_ppc_dist(posterior_predictive_oos_regression_test, kind="ecdf"); |
There was a problem hiding this comment.
might be more distracting than helpful now that the plot is a single line, but we could keep the 0-1000 xlims with
| az.plot_ppc_dist(posterior_predictive_oos_regression_test, kind="ecdf"); | |
| pc = az.plot_ppc_dist(posterior_predictive_oos_regression_test, kind="ecdf") | |
| pc.get_viz("plot", "y").set_xlim(0, 1_000); |
There was a problem hiding this comment.
could be, but also important to show the syntax.
|
|
|
Edit: |
|
I think it is probably better to start the migration before the actual 1.0 release. After that we'll only need to replace Moreover, if we wait, we won't be able to update any page until pymc has fully migrated internally. That means there are many new features like dedicated ppc plots, prior sensitivity checks, loo with moment matching or subsampling, advanced facetting and aesthetic mappings that are ready to use but can't be used because we don't want to use |
3 participants
No description provided.