arviz 0.18.0__py3-none-any.whl → 0.19.0__py3-none-any.whl

arviz/plots/forestplot.py

@@ -246,7 +246,7 @@ def plot_forest(
         width_ratios.append(1)
 
     if hdi_prob is None:
-        hdi_prob = rcParams["stats.hdi_prob"]
+        hdi_prob = rcParams["stats.ci_prob"]
     elif not 1 >= hdi_prob > 0:
         raise ValueError("The value of hdi_prob should be in the interval (0, 1]")
 

arviz/plots/hdiplot.py

@@ -42,7 +42,7 @@ def plot_hdi(
     hdi_data : array_like, optional
         Precomputed HDI values to use. Assumed shape is ``(*x.shape, 2)``.
     hdi_prob : float, optional
-        Probability for the highest density interval. Defaults to ``stats.hdi_prob`` rcParam.
+        Probability for the highest density interval. Defaults to ``stats.ci_prob`` rcParam.
         See :ref:`this section <common_ hdi_prob>` for usage examples.
     color : str, default "C1"
         Color used for the limits of the HDI and fill. Should be a valid matplotlib color.
@@ -155,7 +155,7 @@ def plot_hdi(
     else:
         y = np.asarray(y)
         if hdi_prob is None:
-            hdi_prob = rcParams["stats.hdi_prob"]
+            hdi_prob = rcParams["stats.ci_prob"]
         elif not 1 >= hdi_prob > 0:
             raise ValueError("The value of hdi_prob should be in the interval (0, 1]")
         hdi_data = hdi(y, hdi_prob=hdi_prob, circular=circular, multimodal=False, **hdi_kwargs)

arviz/plots/khatplot.py

@@ -1,6 +1,7 @@
 """Pareto tail indices plot."""
 
 import logging
+import warnings
 
 import numpy as np
 from xarray import DataArray
@@ -40,10 +41,8 @@ def plot_khat(
 
     Parameters
     ----------
-    khats : ELPDData or array-like
-        The input Pareto tail indices to be plotted. It can be an ``ELPDData`` object containing
-        Pareto shapes or an array. In this second case, all the values in the array are interpreted
-        as Pareto tail indices.
+    khats : ELPDData
+        The input Pareto tail indices to be plotted.
     color : str or array_like, default "C0"
         Colors of the scatter plot, if color is a str all dots will have the same color,
         if it is the size of the observations, each dot will have the specified color,
@@ -149,8 +148,9 @@ def plot_khat(
 
     References
     ----------
-    .. [1] Vehtari, A., Simpson, D., Gelman, A., Yao, Y., Gabry, J.,
-        2019. Pareto Smoothed Importance Sampling. arXiv:1507.02646 [stat].
+    .. [1] Vehtari, A., Simpson, D., Gelman, A., Yao, Y., Gabry, J. (2024).
+        Pareto Smoothed Importance Sampling. Journal of Machine Learning
+        Research, 25(72):1-58.
 
     """
     if annotate:
@@ -164,13 +164,29 @@ def plot_khat(
         color = "C0"
 
     if isinstance(khats, np.ndarray):
+        warnings.warn(
+            "support for arrays will be deprecated, please use ELPDData."
+            "The reason for this, is that we need to know the numbers of draws"
+            "sampled from the posterior",
+            FutureWarning,
+        )
         khats = khats.flatten()
         xlabels = False
         legend = False
         dims = []
+        good_k = None
     else:
         if isinstance(khats, ELPDData):
+            good_k = khats.good_k
             khats = khats.pareto_k
+        else:
+            good_k = None
+            warnings.warn(
+                "support for DataArrays will be deprecated, please use ELPDData."
+                "The reason for this, is that we need to know the numbers of draws"
+                "sampled from the posterior",
+                FutureWarning,
+            )
         if not isinstance(khats, DataArray):
             raise ValueError("Incorrect khat data input. Check the documentation")
 
@@ -191,6 +207,7 @@ def plot_khat(
         figsize=figsize,
         xdata=xdata,
         khats=khats,
+        good_k=good_k,
         kwargs=kwargs,
         threshold=threshold,
         coord_labels=coord_labels,

arviz/plots/loopitplot.py

@@ -55,7 +55,7 @@ def plot_loo_pit(
         In this case, instead of overlaying uniform distributions, the beta ``hdi_prob``
         around the theoretical uniform CDF is shown. This approximation only holds
         for large S and ECDF values not very close to 0 nor 1. For more information, see
-        `Vehtari et al. (2019)`, `Appendix G <https://avehtari.github.io/rhat_ess/rhat_ess.html>`_.
+        `Vehtari et al. (2021)`, `Appendix G <https://avehtari.github.io/rhat_ess/rhat_ess.html>`_.
     ecdf_fill : bool, optional
         Use :meth:`matplotlib.axes.Axes.fill_between` to mark the area
         inside the credible interval. Otherwise, plot the
@@ -159,7 +159,7 @@ def plot_loo_pit(
     x_vals = None
 
     if hdi_prob is None:
-        hdi_prob = rcParams["stats.hdi_prob"]
+        hdi_prob = rcParams["stats.ci_prob"]
     elif not 1 >= hdi_prob > 0:
         raise ValueError("The value of hdi_prob should be in the interval (0, 1]")
 

arviz/plots/mcseplot.py

@@ -109,7 +109,9 @@ def plot_mcse(
 
     References
     ----------
-    * Vehtari et al. (2019) see https://arxiv.org/abs/1903.08008
+    .. [1] Vehtari et al. (2021). Rank-normalization, folding, and
+        localization: An improved Rhat for assessing convergence of
+        MCMC. Bayesian analysis, 16(2):667-718.
 
     Examples
     --------

arviz/plots/plot_utils.py

@@ -245,10 +245,8 @@ def format_coords_as_labels(dataarray, skip_dims=None):
     coord_labels = coord_labels.values
     if isinstance(coord_labels[0], tuple):
         fmt = ", ".join(["{}" for _ in coord_labels[0]])
-        coord_labels[:] = [fmt.format(*x) for x in coord_labels]
-    else:
-        coord_labels[:] = [f"{s}" for s in coord_labels]
-    return coord_labels
+        return np.array([fmt.format(*x) for x in coord_labels])
+    return np.array([f"{s}" for s in coord_labels])
 
 
 def set_xticklabels(ax, coord_labels):

arviz/plots/posteriorplot.py

@@ -237,7 +237,7 @@ def plot_posterior(
         labeller = BaseLabeller()
 
     if hdi_prob is None:
-        hdi_prob = rcParams["stats.hdi_prob"]
+        hdi_prob = rcParams["stats.ci_prob"]
     elif hdi_prob not in (None, "hide"):
         if not 1 >= hdi_prob > 0:
             raise ValueError("The value of hdi_prob should be in the interval (0, 1]")

arviz/plots/rankplot.py

@@ -46,8 +46,8 @@ def plot_rank(
     indicates good mixing of the chains.
 
     This plot was introduced by Aki Vehtari, Andrew Gelman, Daniel Simpson, Bob Carpenter,
-    Paul-Christian Burkner (2019): Rank-normalization, folding, and localization: An improved R-hat
-    for assessing convergence of MCMC. arXiv preprint https://arxiv.org/abs/1903.08008
+    Paul-Christian Burkner (2021): Rank-normalization, folding, and localization:
+    An improved R-hat for assessing convergence of MCMC. Bayesian analysis, 16(2):667-718.
 
 
     Parameters

arviz/plots/violinplot.py

@@ -152,7 +152,7 @@ def plot_violin(
     rows, cols = default_grid(len(plotters), grid=grid)
 
     if hdi_prob is None:
-        hdi_prob = rcParams["stats.hdi_prob"]
+        hdi_prob = rcParams["stats.ci_prob"]
     elif not 1 >= hdi_prob > 0:
         raise ValueError("The value of hdi_prob should be in the interval (0, 1]")
 

arviz/preview.py

@@ -0,0 +1,17 @@
+# pylint: disable=unused-import,unused-wildcard-import,wildcard-import
+"""Expose features from arviz-xyz refactored packages inside ``arviz.preview`` namespace."""
+
+try:
+    from arviz_base import *
+except ModuleNotFoundError:
+    pass
+
+try:
+    import arviz_stats
+except ModuleNotFoundError:
+    pass
+
+try:
+    from arviz_plots import *
+except ModuleNotFoundError:
+    pass

arviz/rcparams.py

@@ -26,6 +26,8 @@ _log = logging.getLogger(__name__)
 ScaleKeyword = Literal["log", "negative_log", "deviance"]
 ICKeyword = Literal["loo", "waic"]
 
+_identity = lambda x: x
+
 
 def _make_validate_choice(accepted_values, allow_none=False, typeof=str):
     """Validate value is in accepted_values.
@@ -300,7 +302,7 @@ defaultParams = {  # pylint: disable=invalid-name
         lambda x: x,
     ),
     "plot.matplotlib.show": (False, _validate_boolean),
-    "stats.hdi_prob": (0.94, _validate_probability),
+    "stats.ci_prob": (0.94, _validate_probability),
     "stats.information_criterion": (
         "loo",
         _make_validate_choice({"loo", "waic"} if NO_GET_ARGS else set(get_args(ICKeyword))),
@@ -318,6 +320,9 @@ defaultParams = {  # pylint: disable=invalid-name
     ),
 }
 
+# map from deprecated params to (version, new_param, fold2new, fnew2old)
+deprecated_map = {"stats.hdi_prob": ("0.18.0", "stats.ci_prob", _identity, _identity)}
+
 
 class RcParams(MutableMapping):
     """Class to contain ArviZ default parameters.
@@ -335,6 +340,15 @@ class RcParams(MutableMapping):
 
     def __setitem__(self, key, val):
         """Add validation to __setitem__ function."""
+        if key in deprecated_map:
+            version, key_new, fold2new, _ = deprecated_map[key]
+            warnings.warn(
+                f"{key} is deprecated since {version}, use {key_new} instead",
+                FutureWarning,
+            )
+            key = key_new
+            val = fold2new(val)
+
         try:
             try:
                 cval = self.validate[key](val)
@@ -349,7 +363,18 @@ class RcParams(MutableMapping):
 
     def __getitem__(self, key):
         """Use underlying dict's getitem method."""
-        return self._underlying_storage[key]
+        if key in deprecated_map:
+            version, key_new, _, fnew2old = deprecated_map[key]
+            warnings.warn(
+                f"{key} is deprecated since {version}, use {key_new} instead",
+                FutureWarning,
+            )
+            if key not in self._underlying_storage:
+                key = key_new
+        else:
+            fnew2old = _identity
+
+        return fnew2old(self._underlying_storage[key])
 
     def __delitem__(self, key):
         """Raise TypeError if someone ever tries to delete a key from RcParams."""

arviz/stats/diagnostics.py

@@ -135,10 +135,11 @@ def ess(
 
     References
     ----------
-    * Vehtari et al. (2019) see https://arxiv.org/abs/1903.08008
-    * https://mc-stan.org/docs/2_18/reference-manual/effective-sample-size-section.html
-      Section 15.4.2
-    * Gelman et al. BDA (2014) Formula 11.8
+    * Vehtari et al. (2021). Rank-normalization, folding, and
+        localization: An improved Rhat for assessing convergence of
+        MCMC. Bayesian analysis, 16(2):667-718.
+    * https://mc-stan.org/docs/reference-manual/analysis.html#effective-sample-size.section
+    * Gelman et al. BDA3 (2013) Formula 11.8
 
     See Also
     --------
@@ -246,7 +247,7 @@ def rhat(data, *, var_names=None, method="rank", dask_kwargs=None):
         Names of variables to include in the rhat report
     method : str
         Select R-hat method. Valid methods are:
-        - "rank"        # recommended by Vehtari et al. (2019)
+        - "rank"        # recommended by Vehtari et al. (2021)
         - "split"
         - "folded"
         - "z_scale"
@@ -269,7 +270,7 @@ def rhat(data, *, var_names=None, method="rank", dask_kwargs=None):
     -----
     The diagnostic is computed by:
 
-      .. math:: \hat{R} = \frac{\hat{V}}{W}
+      .. math:: \hat{R} = \sqrt{\frac{\hat{V}}{W}}
 
     where :math:`W` is the within-chain variance and :math:`\hat{V}` is the posterior variance
     estimate for the pooled rank-traces. This is the potential scale reduction factor, which
@@ -277,12 +278,15 @@ def rhat(data, *, var_names=None, method="rank", dask_kwargs=None):
     greater than one indicate that one or more chains have not yet converged.
 
     Rank values are calculated over all the chains with ``scipy.stats.rankdata``.
-    Each chain is split in two and normalized with the z-transform following Vehtari et al. (2019).
+    Each chain is split in two and normalized with the z-transform following
+    Vehtari et al. (2021).
 
     References
     ----------
-    * Vehtari et al. (2019) see https://arxiv.org/abs/1903.08008
-    * Gelman et al. BDA (2014)
+    * Vehtari et al. (2021).  Rank-normalization, folding, and
+      localization: An improved Rhat for assessing convergence of
+      MCMC. Bayesian analysis, 16(2):667-718.
+    * Gelman et al. BDA3 (2013)
     * Brooks and Gelman (1998)
     * Gelman and Rubin (1992)
 

arviz/stats/ecdf_utils.py

@@ -25,6 +25,13 @@ def _get_ecdf_points(
     return x, y
 
 
+def _call_rvs(rvs, ndraws, random_state):
+    if random_state is None:
+        return rvs(ndraws)
+    else:
+        return rvs(ndraws, random_state=random_state)
+
+
 def _simulate_ecdf(
     ndraws: int,
     eval_points: np.ndarray,
@@ -32,7 +39,7 @@ def _simulate_ecdf(
     random_state: Optional[Any] = None,
 ) -> np.ndarray:
     """Simulate ECDF at the `eval_points` using the given random variable sampler"""
-    sample = rvs(ndraws, random_state=random_state)
+    sample = _call_rvs(rvs, ndraws, random_state)
     sample.sort()
     return compute_ecdf(sample, eval_points)
 
@@ -91,14 +98,10 @@ def ecdf_confidence_band(
         A function that takes an integer `ndraws` and optionally the object passed to
         `random_state` and returns an array of `ndraws` samples from the same distribution
         as the original dataset. Required if `method` is "simulated" and variable is discrete.
-    num_trials : int, default 1000
+    num_trials : int, default 500
         The number of random ECDFs to generate for constructing simultaneous confidence bands
         (if `method` is "simulated").
-    random_state : {None, int, `numpy.random.Generator`,
-                    `numpy.random.RandomState`}, optional
-        If `None`, the `numpy.random.RandomState` singleton is used. If an `int`, a new
-        ``numpy.random.RandomState`` instance is used, seeded with seed. If a `RandomState` or
-        `Generator` instance, the instance is used.
+    random_state : int, numpy.random.Generator or numpy.random.RandomState, optional
 
     Returns
     -------
@@ -132,7 +135,7 @@ def _simulate_simultaneous_ecdf_band_probability(
     cdf_at_eval_points: np.ndarray,
     prob: float = 0.95,
     rvs: Optional[Callable[[int, Optional[Any]], np.ndarray]] = None,
-    num_trials: int = 1000,
+    num_trials: int = 500,
     random_state: Optional[Any] = None,
 ) -> float:
     """Estimate probability for simultaneous confidence band using simulation.

arviz/stats/stats.py

@@ -270,12 +270,12 @@ def compare(
             weights[i] = u_weights / np.sum(u_weights)
 
         weights = weights.mean(axis=0)
-        ses = pd.Series(z_bs.std(axis=0), index=names)  # pylint: disable=no-member
+        ses = pd.Series(z_bs.std(axis=0), index=ics.index)  # pylint: disable=no-member
 
     elif method.lower() == "pseudo-bma":
         min_ic = ics.iloc[0][f"elpd_{ic}"]
         z_rv = np.exp((ics[f"elpd_{ic}"] - min_ic) / scale_value)
-        weights = z_rv / np.sum(z_rv)
+        weights = (z_rv / np.sum(z_rv)).to_numpy()
         ses = ics["se"]
 
     if np.any(weights):
@@ -471,7 +471,7 @@ def hdi(
         Refer to documentation of :func:`arviz.convert_to_dataset` for details.
     hdi_prob: float, optional
         Prob for which the highest density interval will be computed. Defaults to
-        ``stats.hdi_prob`` rcParam.
+        ``stats.ci_prob`` rcParam.
     circular: bool, optional
         Whether to compute the hdi taking into account `x` is a circular variable
         (in the range [-np.pi, np.pi]) or not. Defaults to False (i.e non-circular variables).
@@ -553,7 +553,7 @@ def hdi(
 
     """
     if hdi_prob is None:
-        hdi_prob = rcParams["stats.hdi_prob"]
+        hdi_prob = rcParams["stats.ci_prob"]
     elif not 1 >= hdi_prob > 0:
         raise ValueError("The value of hdi_prob should be in the interval (0, 1]")
 
@@ -715,8 +715,9 @@ def loo(data, pointwise=None, var_name=None, reff=None, scale=None):
     se: standard error of the elpd
     p_loo: effective number of parameters
     shape_warn: bool
-        True if the estimated shape parameter of
-        Pareto distribution is greater than 0.7 for one or more samples
+        True if the estimated shape parameter of Pareto distribution is greater than a thresold
+        value  for one or more samples. For a sample size S, the thresold is compute as
+        min(1 - 1/log10(S), 0.7)
     loo_i: array of pointwise predictive accuracy, only if pointwise True
     pareto_k: array of Pareto shape values, only if pointwise True
     scale: scale of the elpd
@@ -785,13 +786,15 @@ def loo(data, pointwise=None, var_name=None, reff=None, scale=None):
     log_weights += log_likelihood
 
     warn_mg = False
-    if np.any(pareto_shape > 0.7):
+    good_k = min(1 - 1 / np.log10(n_samples), 0.7)
+
+    if np.any(pareto_shape > good_k):
         warnings.warn(
-            "Estimated shape parameter of Pareto distribution is greater than 0.7 for "
-            "one or more samples. You should consider using a more robust model, this is because "
-            "importance sampling is less likely to work well if the marginal posterior and "
-            "LOO posterior are very different. This is more likely to happen with a non-robust "
-            "model and highly influential observations."
+            f"Estimated shape parameter of Pareto distribution is greater than {good_k:.2f} "
+            "for one or more samples. You should consider using a more robust model, this is "
+            "because importance sampling is less likely to work well if the marginal posterior "
+            "and LOO posterior are very different. This is more likely to happen with a "
+            "non-robust model and highly influential observations."
         )
         warn_mg = True
 
@@ -816,8 +819,17 @@ def loo(data, pointwise=None, var_name=None, reff=None, scale=None):
 
     if not pointwise:
         return ELPDData(
-            data=[loo_lppd, loo_lppd_se, p_loo, n_samples, n_data_points, warn_mg, scale],
-            index=["elpd_loo", "se", "p_loo", "n_samples", "n_data_points", "warning", "scale"],
+            data=[loo_lppd, loo_lppd_se, p_loo, n_samples, n_data_points, warn_mg, scale, good_k],
+            index=[
+                "elpd_loo",
+                "se",
+                "p_loo",
+                "n_samples",
+                "n_data_points",
+                "warning",
+                "scale",
+                "good_k",
+            ],
         )
     if np.equal(loo_lppd, loo_lppd_i).all():  # pylint: disable=no-member
         warnings.warn(
@@ -835,6 +847,7 @@ def loo(data, pointwise=None, var_name=None, reff=None, scale=None):
             loo_lppd_i.rename("loo_i"),
             pareto_shape,
             scale,
+            good_k,
         ],
         index=[
             "elpd_loo",
@@ -846,6 +859,7 @@ def loo(data, pointwise=None, var_name=None, reff=None, scale=None):
             "loo_i",
             "pareto_k",
             "scale",
+            "good_k",
         ],
     )
 
@@ -879,7 +893,8 @@ def psislw(log_weights, reff=1.0):
 
     References
     ----------
-    * Vehtari et al. (2015) see https://arxiv.org/abs/1507.02646
+    * Vehtari et al. (2024). Pareto smoothed importance sampling. Journal of Machine
+      Learning Research, 25(72):1-58.
 
     See Also
     --------
@@ -1322,7 +1337,7 @@ def summary(
     if labeller is None:
         labeller = BaseLabeller()
     if hdi_prob is None:
-        hdi_prob = rcParams["stats.hdi_prob"]
+        hdi_prob = rcParams["stats.ci_prob"]
     elif not 1 >= hdi_prob > 0:
         raise ValueError("The value of hdi_prob should be in the interval (0, 1]")
 

arviz/stats/stats_utils.py

@@ -454,10 +454,9 @@ POINTWISE_LOO_FMT = """------
 
 Pareto k diagnostic values:
                          {{0:>{0}}} {{1:>6}}
-(-Inf, 0.5]   (good)     {{2:{0}d}} {{6:6.1f}}%
- (0.5, 0.7]   (ok)       {{3:{0}d}} {{7:6.1f}}%
-   (0.7, 1]   (bad)      {{4:{0}d}} {{8:6.1f}}%
-   (1, Inf)   (very bad) {{5:{0}d}} {{9:6.1f}}%
+(-Inf, {{8:.2f}}]   (good)     {{2:{0}d}} {{5:6.1f}}%
+   ({{8:.2f}}, 1]   (bad)      {{3:{0}d}} {{6:6.1f}}%
+   (1, Inf)   (very bad) {{4:{0}d}} {{7:6.1f}}%
 """
 SCALE_DICT = {"deviance": "deviance", "log": "elpd", "negative_log": "-elpd"}
 
@@ -488,11 +487,14 @@ class ELPDData(pd.Series):  # pylint: disable=too-many-ancestors
             base += "\n\nThere has been a warning during the calculation. Please check the results."
 
         if kind == "loo" and "pareto_k" in self:
-            bins = np.asarray([-np.inf, 0.5, 0.7, 1, np.inf])
+            bins = np.asarray([-np.inf, self.good_k, 1, np.inf])
             counts, *_ = _histogram(self.pareto_k.values, bins)
             extended = POINTWISE_LOO_FMT.format(max(4, len(str(np.max(counts)))))
             extended = extended.format(
-                "Count", "Pct.", *[*counts, *(counts / np.sum(counts) * 100)]
+                "Count",
+                "Pct.",
+                *[*counts, *(counts / np.sum(counts) * 100)],
+                self.good_k,
             )
             base = "\n".join([base, extended])
         return base

arviz/tests/base_tests/test_data.py

@@ -42,7 +42,6 @@ from ..helpers import (  # pylint: disable=unused-import
     draws,
     eight_schools_params,
     models,
-    running_on_ci,
 )
 
 
@@ -1469,7 +1468,7 @@ class TestJSON:
 
 
 @pytest.mark.skipif(
-    not (importlib.util.find_spec("datatree") or running_on_ci()),
+    not (importlib.util.find_spec("datatree") or "ARVIZ_REQUIRE_ALL_DEPS" in os.environ),
     reason="test requires xarray-datatree library",
 )
 class TestDataTree:

arviz/tests/base_tests/test_data_zarr.py

@@ -16,7 +16,6 @@ from ..helpers import (  # pylint: disable=unused-import
     draws,
     eight_schools_params,
     importorskip,
-    running_on_ci,
 )
 
 zarr = importorskip("zarr")  # pylint: disable=invalid-name

arviz/tests/base_tests/test_diagnostics_numba.py

@@ -1,7 +1,5 @@
 """Test Diagnostic methods"""
 
-import importlib
-
 # pylint: disable=redefined-outer-name, no-member, too-many-public-methods
 import numpy as np
 import pytest
@@ -11,13 +9,10 @@ from ...rcparams import rcParams
 from ...stats import bfmi, mcse, rhat
 from ...stats.diagnostics import _mc_error, ks_summary
 from ...utils import Numba
-from ..helpers import running_on_ci
+from ..helpers import importorskip
 from .test_diagnostics import data  # pylint: disable=unused-import
 
-pytestmark = pytest.mark.skipif(  # pylint: disable=invalid-name
-    (importlib.util.find_spec("numba") is None) and not running_on_ci(),
-    reason="test requires numba which is not installed",
-)
+importorskip("numba")
 
 rcParams["data.load"] = "eager"
 

arviz/tests/base_tests/test_helpers.py

@@ -6,13 +6,13 @@ from ..helpers import importorskip
 
 def test_importorskip_local(monkeypatch):
     """Test ``importorskip`` run on local machine with non-existent module, which should skip."""
-    monkeypatch.delenv("ARVIZ_CI_MACHINE", raising=False)
+    monkeypatch.delenv("ARVIZ_REQUIRE_ALL_DEPS", raising=False)
     with pytest.raises(Skipped):
         importorskip("non-existent-function")
 
 
 def test_importorskip_ci(monkeypatch):
     """Test ``importorskip`` run on CI machine with non-existent module, which should fail."""
-    monkeypatch.setenv("ARVIZ_CI_MACHINE", 1)
+    monkeypatch.setenv("ARVIZ_REQUIRE_ALL_DEPS", 1)
     with pytest.raises(ModuleNotFoundError):
         importorskip("non-existent-function")

arviz/tests/base_tests/test_plot_utils.py

@@ -1,5 +1,6 @@
 # pylint: disable=redefined-outer-name
 import importlib
+import os
 
 import numpy as np
 import pytest
@@ -20,10 +21,10 @@ from ...rcparams import rc_context
 from ...sel_utils import xarray_sel_iter, xarray_to_ndarray
 from ...stats.density_utils import get_bins
 from ...utils import get_coords
-from ..helpers import running_on_ci
 
 # Check if Bokeh is installed
 bokeh_installed = importlib.util.find_spec("bokeh") is not None  # pylint: disable=invalid-name
+skip_tests = (not bokeh_installed) and ("ARVIZ_REQUIRE_ALL_DEPS" not in os.environ)
 
 
 @pytest.mark.parametrize(
@@ -212,10 +213,7 @@ def test_filter_plotter_list_warning():
     assert len(plotters_filtered) == 5
 
 
-@pytest.mark.skipif(
-    not (bokeh_installed or running_on_ci()),
-    reason="test requires bokeh which is not installed",
-)
+@pytest.mark.skipif(skip_tests, reason="test requires bokeh which is not installed")
 def test_bokeh_import():
     """Tests that correct method is returned on bokeh import"""
     plot = get_plotting_function("plot_dist", "distplot", "bokeh")
@@ -290,10 +288,7 @@ def test_mpl_dealiase_sel_kwargs():
     assert res["line_color"] == "red"
 
 
-@pytest.mark.skipif(
-    not (bokeh_installed or running_on_ci()),
-    reason="test requires bokeh which is not installed",
-)
+@pytest.mark.skipif(skip_tests, reason="test requires bokeh which is not installed")
 def test_bokeh_dealiase_sel_kwargs():
     """Check bokeh dealiase_sel_kwargs behaviour.
 
@@ -315,10 +310,7 @@ def test_bokeh_dealiase_sel_kwargs():
     assert res["line_color"] == "red"
 
 
-@pytest.mark.skipif(
-    not (bokeh_installed or running_on_ci()),
-    reason="test requires bokeh which is not installed",
-)
+@pytest.mark.skipif(skip_tests, reason="test requires bokeh which is not installed")
 def test_set_bokeh_circular_ticks_labels():
     """Assert the axes returned after placing ticks and tick labels for circular plots."""
     import bokeh.plotting as bkp