Add panel covariates to plot_cap() and make it more flexible

[tomicapretto]

This is still WIP, but I'm excited about this PR since it makes plot_cap() much more flexible

Now we can do something like

data = pd.read_csv("https://gist.githubusercontent.com/seankross/a412dfbd88b3db70b74b/raw/5f23f993cd87c283ce766e7ac6b329ee7cc2e1d1/mtcars.csv")
data["cyl"] = data["cyl"].replace({4: "low", 6: "medium", 8: "high"})
data["cyl"] = pd.Categorical(data["cyl"], categories=["low", "medium", "high"], ordered=True)
data["gear"] = data["gear"].replace({3: "A", 4: "B", 5: "C"})
model = bmb.Model("mpg ~ 0 + hp * wt + cyl + gear", data)
idata = model.fit(draws=500, target_accept=0.9, random_seed=1234)

fig, axes = plot_cap(
    model, 
    idata, 
    {"horizontal": "hp", "color": "cyl","panel": "gear"},
    use_hdi=False, 
    fig_kwargs={"figsize": (15, 5), "dpi": 120, "sharey": True}
);

All the previous examples still work. I still need to update the the internal function _plot_cap_categoric() and clean some details in _plot_cap_numeric(). In particular I need to handle legends using the figure and not the axes.

Edit: This is done. I decided to create a notebook to show the new features of plot_cap(). Basically, it still works for all the cases it used to work, but now it's more flexible. We can map groups to panels.

https://gist.github.com/tomicapretto/1388437357d8520ff5ff2fcee39b7d1b

[codecov-commenter]

Codecov Report

Merging #596 (1291406) into main (3af8328) will decrease coverage by 2.49%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##             main     #596      +/-   ##
==========================================
- Coverage   85.36%   82.86%   -2.50%     
==========================================
  Files          38       38              
  Lines        2944     3034      +90     
==========================================
+ Hits         2513     2514       +1     
- Misses        431      520      +89     

Impacted Files	Coverage Δ
bambi/plots/plot_cap.py	0.00% <0.00%> (ø)
bambi/tests/test_built_models.py	96.50% <0.00%> (+0.29%)	⬆️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[canyon289]

Suggested change

	name of the variables, and the values are functions to be applied. Defaults to ``None``.
	name of the variables, and the values are functions to be applied. Defaults to `None`.

[canyon289]

Actually, just curious why are the double ticks needed? I see it in other docstrings too.

[tomicapretto]

If I remember correctly it's because it doesn't render properly if it only has a single backtick. Maybe something changed and now it works with a single backtick, but previously I remember trying a single backtick without working

[canyon289]

This is incredible, you basically rewrote this whole module, and now it can do so much more

[canyon289]

you're probably right just ignore me

…

On Sunday, November 27, 2022, Tomás Capretto ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In bambi/plots/plot_cap.py <#596 (comment)>: > @@ -133,9 +150,12 @@ def plot_cap(model, idata, covariates, use_hdi=True, hdi_prob=None, legend=True, Changing the global variable ``az.rcParam["stats.hdi_prob"]`` affects this default. legend : bool, optional Whether to automatically include a legend in the plot. Defaults to ``True``. + transforms : dict, optional + Transformations that are applied to each of the variables being plotted. The keys are the + name of the variables, and the values are functions to be applied. Defaults to ``None``. If I remember correctly it's because it doesn't render properly if it only has a single backtick. Maybe something changed and now it works with a single backtick, but previously I remember trying a single backtick without working — Reply to this email directly, view it on GitHub <#596 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABXBFYJ3T3W5WGK3BRAMM4TWKNRPPANCNFSM6AAAAAASMBVQKA> . You are receiving this because you commented.Message ID: ***@***.***>