bayesian-pricing 0.2.3__tar.gz → 0.2.4__tar.gz

{bayesian_pricing-0.2.3 → bayesian_pricing-0.2.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: bayesian-pricing
-Version: 0.2.3
+Version: 0.2.4
 Summary: Hierarchical Bayesian models for insurance pricing thin-data segments
 Project-URL: Homepage, https://github.com/burning-cost/bayesian-pricing
 Project-URL: Repository, https://github.com/burning-cost/bayesian-pricing
@@ -42,7 +42,7 @@ Description-Content-Type: text/markdown
 [![PyPI](https://img.shields.io/pypi/v/bayesian-pricing)](https://pypi.org/project/bayesian-pricing/)
 [![Python](https://img.shields.io/pypi/pyversions/bayesian-pricing)](https://pypi.org/project/bayesian-pricing/)
 [![Tests](https://img.shields.io/badge/tests-passing-brightgreen)]()
-[![License](https://img.shields.io/badge/license-BSD--3-blue)]()
+[![License](https://img.shields.io/badge/license-MIT-green)]()
 
 Hierarchical Bayesian models for insurance pricing thin-data segments — when your rating grid has more cells than your book has claims, partial pooling gives you credible estimates where every other method gives you noise or nothing.
 
@@ -138,6 +138,7 @@ model.fit(df, claim_count_col="claims", exposure_col="exposure", sampler_config=
 # Posterior predictive means for each segment
 preds = model.predict()
 print(preds)
+# Output is illustrative — exact values depend on your data and sampler seed:
 #   veh_group age_band      mean      p5       p50      p95  credibility_factor
 #   Supermini    17-21    0.1234  0.0812   0.1201   0.1731           0.38
 #   Sports       17-21    0.1891  0.1102   0.1845   0.2881           0.21  <- thin
@@ -158,19 +159,19 @@ tables = rel.relativities()
 # Single factor in rate-table format
 veh_table = rel.relativities(factor="veh_group")
 print(veh_table.table)
-#   level      relativity  lower_90pct  upper_90pct  credibility_factor  interval_width
-#   Sports          1.524        1.234        1.891           0.71               0.657
-#   Saloon          1.000        0.921        1.082           0.94               0.161
-#   Supermini       0.819        0.764        0.881           0.89               0.117
+#   level      relativity  lower_90pct  upper_90pct  uncertainty_reduction  interval_width
+#   Sports          1.524        1.234        1.891                 0.71               0.657
+#   Saloon          1.000        0.921        1.082                 0.94               0.161
+#   Supermini       0.819        0.764        0.881                 0.89               0.117
 
 # Identify thin segments that need manual review
 thin = rel.thin_segments(credibility_threshold=0.3)
 print(thin)
-# factor    level    credibility_factor    relativity
-# veh_group Sports-17-21          0.18         1.84   <- sparse cell, wide CI
+# factor    level    uncertainty_reduction    relativity
+# veh_group Sports-17-21          0.18          1.84   <- sparse cell, wide CI
 
 # Export for Excel / rate system import
-summary_df = rel.summary()  # long format: factor, level, relativity, CI, credibility
+summary_df = rel.summary()  # long format: factor, level, relativity, CI, uncertainty_reduction
 summary_df.write_csv("bayesian_relativities.csv")
 ```
 

{bayesian_pricing-0.2.3 → bayesian_pricing-0.2.4}/README.md

@@ -3,7 +3,7 @@
 [![PyPI](https://img.shields.io/pypi/v/bayesian-pricing)](https://pypi.org/project/bayesian-pricing/)
 [![Python](https://img.shields.io/pypi/pyversions/bayesian-pricing)](https://pypi.org/project/bayesian-pricing/)
 [![Tests](https://img.shields.io/badge/tests-passing-brightgreen)]()
-[![License](https://img.shields.io/badge/license-BSD--3-blue)]()
+[![License](https://img.shields.io/badge/license-MIT-green)]()
 
 Hierarchical Bayesian models for insurance pricing thin-data segments — when your rating grid has more cells than your book has claims, partial pooling gives you credible estimates where every other method gives you noise or nothing.
 
@@ -99,6 +99,7 @@ model.fit(df, claim_count_col="claims", exposure_col="exposure", sampler_config=
 # Posterior predictive means for each segment
 preds = model.predict()
 print(preds)
+# Output is illustrative — exact values depend on your data and sampler seed:
 #   veh_group age_band      mean      p5       p50      p95  credibility_factor
 #   Supermini    17-21    0.1234  0.0812   0.1201   0.1731           0.38
 #   Sports       17-21    0.1891  0.1102   0.1845   0.2881           0.21  <- thin
@@ -119,19 +120,19 @@ tables = rel.relativities()
 # Single factor in rate-table format
 veh_table = rel.relativities(factor="veh_group")
 print(veh_table.table)
-#   level      relativity  lower_90pct  upper_90pct  credibility_factor  interval_width
-#   Sports          1.524        1.234        1.891           0.71               0.657
-#   Saloon          1.000        0.921        1.082           0.94               0.161
-#   Supermini       0.819        0.764        0.881           0.89               0.117
+#   level      relativity  lower_90pct  upper_90pct  uncertainty_reduction  interval_width
+#   Sports          1.524        1.234        1.891                 0.71               0.657
+#   Saloon          1.000        0.921        1.082                 0.94               0.161
+#   Supermini       0.819        0.764        0.881                 0.89               0.117
 
 # Identify thin segments that need manual review
 thin = rel.thin_segments(credibility_threshold=0.3)
 print(thin)
-# factor    level    credibility_factor    relativity
-# veh_group Sports-17-21          0.18         1.84   <- sparse cell, wide CI
+# factor    level    uncertainty_reduction    relativity
+# veh_group Sports-17-21          0.18          1.84   <- sparse cell, wide CI
 
 # Export for Excel / rate system import
-summary_df = rel.summary()  # long format: factor, level, relativity, CI, credibility
+summary_df = rel.summary()  # long format: factor, level, relativity, CI, uncertainty_reduction
 summary_df.write_csv("bayesian_relativities.csv")
 ```
 

{bayesian_pricing-0.2.3 → bayesian_pricing-0.2.4}/notebooks/01_hierarchical_frequency_demo.py

@@ -352,7 +352,7 @@ veh_table["recovery_error_pct"] = (
 )
 print("Vehicle group relativity recovery:")
 display(veh_table[["level", "true_relativity", "relativity", "lower_90pct", "upper_90pct",
-                    "recovery_error_pct", "credibility_factor"]].round(3))
+                    "recovery_error_pct", "uncertainty_reduction"]].round(3))
 
 # COMMAND ----------
 
@@ -554,8 +554,8 @@ display(sev_rt.table)
 freq_rt = rel.relativities(factor="veh_group")
 
 # Merge frequency and severity relativities
-pp_table = freq_rt.table[["level", "relativity", "credibility_factor"]].merge(
-    sev_rt.table[["level", "relativity", "credibility_factor"]],
+pp_table = freq_rt.table[["level", "relativity", "uncertainty_reduction"]].rename({"uncertainty_reduction": "uncertainty_reduction_freq"}).merge(
+    sev_rt.table[["level", "relativity", "uncertainty_reduction"]].rename({"uncertainty_reduction": "uncertainty_reduction_sev"}),
     on="level",
     suffixes=("_freq", "_sev"),
 )

{bayesian_pricing-0.2.3 → bayesian_pricing-0.2.4}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "bayesian-pricing"
-version = "0.2.3"
+version = "0.2.4"
 description = "Hierarchical Bayesian models for insurance pricing thin-data segments"
 readme = "README.md"
 requires-python = ">=3.10"

{bayesian_pricing-0.2.3 → bayesian_pricing-0.2.4}/src/bayesian_pricing/__init__.py

@@ -37,19 +37,25 @@ Usage::
 
     rel = BayesianRelativities(freq_model)
     print(rel.relativities())          # Polars DataFrame with posterior median + credible interval
-    print(rel.credibility_factors())   # How much weight each segment puts on own data
+    print(rel.credibility_factors())   # Uncertainty reduction per segment (how data-dominated each level is)
 """
 
-from bayesian_pricing.frequency import HierarchicalFrequency
+from bayesian_pricing.frequency import HierarchicalFrequency, SamplerConfig
 from bayesian_pricing.severity import HierarchicalSeverity
 from bayesian_pricing.relativities import BayesianRelativities
 from bayesian_pricing.diagnostics import convergence_summary, posterior_predictive_check
 
-__version__ = "0.2.1"
+from importlib.metadata import version, PackageNotFoundError
+
+try:
+    __version__ = version("bayesian-pricing")
+except PackageNotFoundError:
+    __version__ = "0.0.0"  # not installed
 __all__ = [
     "HierarchicalFrequency",
     "HierarchicalSeverity",
     "BayesianRelativities",
+    "SamplerConfig",
     "convergence_summary",
     "posterior_predictive_check",
 ]

{bayesian_pricing-0.2.3 → bayesian_pricing-0.2.4}/src/bayesian_pricing/frequency.py

@@ -236,6 +236,15 @@ class HierarchicalFrequency:
         _validate_columns_present(df, self.group_cols + [claim_count_col, exposure_col])
         _validate_positive(df[exposure_col], exposure_col)
 
+        # Validate interaction pairs before _check_pymc() so users get
+        # ValueError rather than ImportError for bad pair specifications.
+        for col_a, col_b in self.interaction_pairs:
+            if col_a not in self.group_cols or col_b not in self.group_cols:
+                raise ValueError(
+                    f"Interaction pair ({col_a!r}, {col_b!r}) references a column "
+                    f"not in group_cols: {self.group_cols}"
+                )
+
         _check_pymc()
         import pymc as pm
 
@@ -259,15 +268,6 @@ class HierarchicalFrequency:
             group_indices[col] = idx
             group_levels[col] = levels
 
-        # Build interaction indices
-        interaction_indices: dict[tuple[str, str], np.ndarray] = {}
-        for col_a, col_b in self.interaction_pairs:
-            if col_a not in self.group_cols or col_b not in self.group_cols:
-                raise ValueError(
-                    f"Interaction pair ({col_a!r}, {col_b!r}) references a column "
-                    f"not in group_cols: {self.group_cols}"
-                )
-
         # Coords for PyMC named dimensions
         coords = {}
         for col in self.group_cols:
@@ -421,8 +421,10 @@ class HierarchicalFrequency:
             - The group columns from your training data
             - mean: posterior mean claim rate (use this as your point estimate)
             - The quantiles you specified (default: p5, p50, p95)
-            - credibility_factor: how much this segment was pulled toward the portfolio
-              mean (0 = fully pooled to mean, 1 = trusts own data completely)
+            - posterior_shrinkage_ratio: measures how far the posterior mean sits between
+              the portfolio mean (0) and the raw observed rate (1). This is NOT the
+              Bühlmann-Straub credibility factor Z; it is a post-hoc ratio. Use
+              variance_components() for the B-S-analogous variance decomposition.
 
         Args:
             quantiles: Posterior quantiles to return. Default gives 90% credible
@@ -468,15 +470,23 @@ class HierarchicalFrequency:
         for q in quantiles:
             result_dict[f"p{int(q * 100)}"] = np.quantile(rate_samples, q, axis=0).tolist()
 
-        # Credibility factor: Z = (posterior mean - portfolio mean) / (observed rate - portfolio mean)
-        # Undefined when observed rate == portfolio mean; clamp to [0, 1].
+        # posterior_shrinkage_ratio: how far the posterior mean sits between the
+        # portfolio mean (0) and the raw observed rate (1).
+        #
+        # This is NOT the Bühlmann-Straub credibility factor Z. The B-S Z is derived
+        # analytically as n_i / (n_i + k) where k = sigma^2 / tau^2 (ratio of
+        # within-segment variance to between-segment variance). This column instead
+        # measures posterior shrinkage post-hoc:
+        #     (posterior_mean - portfolio_mean) / (observed_rate - portfolio_mean)
+        # It is clipped to [0, 1] and set to 0.5 when the denominator is near zero.
+        # Use variance_components() for the B-S-analogous tau^2 estimates.
         portfolio_mean = float(np.exp(alpha_samples.mean()))
         obs = self._get_observed_rates()
         denom = obs - portfolio_mean
         num = np.array(result_dict["mean"]) - portfolio_mean
         with np.errstate(divide="ignore", invalid="ignore"):
             z = np.where(np.abs(denom) < 1e-10, 0.5, np.clip(num / denom, 0, 1))
-        result_dict["credibility_factor"] = z.tolist()
+        result_dict["posterior_shrinkage_ratio"] = z.tolist()
 
         return pl.DataFrame(result_dict)
 
@@ -510,7 +520,11 @@ class HierarchicalFrequency:
             hdi_prob=0.94,
         )
 
-        # Add human interpretation column
+        # Add human interpretation column.
+        # typical_relativity_spread = exp(sigma) - 1 is the approximate coefficient
+        # of variation of the lognormal random effect: for a segment at mu + sigma,
+        # the relativity is exp(sigma), so exp(sigma) - 1 is the fractional spread
+        # above 1.0. Provides an intuitive %-spread for reviewers.
         summary_pd["typical_relativity_spread"] = np.exp(summary_pd["mean"]) - 1
 
         # Convert to Polars, bringing the index (parameter names) in as a column

{bayesian_pricing-0.2.3 → bayesian_pricing-0.2.4}/src/bayesian_pricing/relativities.py

@@ -44,14 +44,16 @@ class RelativityTable:
         factor: The grouping column name (e.g., "veh_group").
         levels: List of levels in the factor.
         table: Polars DataFrame with columns: level, relativity, lower, upper,
-            credibility_factor, interval_width.
+            uncertainty_reduction, interval_width.
 
-    The credibility_factor tells you how data-dominated each level is:
-        0.0 = fully pooled to portfolio mean (estimate is 1.0 relativity)
-        1.0 = trusts own experience completely (like a fixed-effects GLM)
-        0.6 = weighted blend: 60% own data, 40% portfolio mean
+    The uncertainty_reduction column measures how much the posterior has reduced
+    uncertainty relative to the prior: 1 - posterior_std / prior_rel_std. A value
+    near 1.0 means the data has largely eliminated prior uncertainty (thick segment).
+    A value near 0.0 means posterior uncertainty is still close to prior uncertainty
+    (thin segment, minimal data influence). This is distinct from the B-S credibility
+    factor Z -- it measures uncertainty reduction, not the mixing weight on own data.
 
-    Levels with credibility_factor < 0.3 have wide credible intervals and
+    Levels with uncertainty_reduction < 0.3 have wide credible intervals and
     should be treated with caution. They should not drive large rate changes.
     """
 
@@ -162,8 +164,11 @@ class BayesianRelativities:
         point_estimate = np.median(rel_samples, axis=0)
         hdi_lower, hdi_upper = self._compute_hdi(rel_samples)
 
-        # Credibility factor heuristic: compare posterior uncertainty to prior uncertainty.
-        # Low posterior std relative to prior std -> high credibility (less uncertain).
+        # Uncertainty reduction: compare posterior uncertainty to prior uncertainty.
+        # 1 - posterior_std / prior_rel_std. Near 1 = data has resolved uncertainty (thick
+        # segment). Near 0 = posterior still similar to prior (thin, heavily pooled segment).
+        # Note: this is NOT the same as the B-S credibility Z from predict(), which measures
+        # the shrinkage fraction (posterior_mean - portfolio_mean) / (observed - portfolio_mean).
         prior_std = np.ones(len(levels)) * self._model.variance_prior_sigma
         posterior_std = rel_samples.std(axis=0)
         prior_rel_std = np.exp(prior_std) - 1
@@ -177,21 +182,26 @@ class BayesianRelativities:
             "relativity": point_estimate.tolist(),
             lower_col: hdi_lower.tolist(),
             upper_col: hdi_upper.tolist(),
-            "credibility_factor": credibility.tolist(),
+            "uncertainty_reduction": credibility.tolist(),
             "interval_width": (hdi_upper - hdi_lower).tolist(),
         }).sort("relativity", descending=True)
 
         return RelativityTable(factor=col, levels=levels, table=table)
 
     def credibility_factors(self) -> "pl.DataFrame":
-        """Return credibility factors for all segments across all factors.
+        """Return uncertainty_reduction for all segments across all factors.
 
-        The credibility factor answers: "what fraction of this segment's estimate
-        comes from its own experience vs the portfolio mean?" This is the Bayesian
-        equivalent of Z_j = w_j / (w_j + K) in Bühlmann-Straub.
+        The uncertainty_reduction column measures how much the posterior has reduced
+        uncertainty relative to the prior: 1 - posterior_std / prior_rel_std.
+        A value near 1.0 means the data has largely eliminated prior uncertainty.
+        A value near 0.0 means the segment is thin and heavily pooled.
+
+        Note: this is distinct from the B-S credibility factor Z returned by
+        predict(). Use predict() credibility_factor to understand shrinkage;
+        use this method to understand how data-dominated each segment is.
 
         Returns:
-            Polars DataFrame with factor, level, credibility_factor, and relativity columns.
+            Polars DataFrame with factor, level, uncertainty_reduction, and relativity columns.
         """
         import polars as pl
 
@@ -202,7 +212,7 @@ class BayesianRelativities:
                 rows.append({
                     "factor": col,
                     "level": row["level"],
-                    "credibility_factor": row["credibility_factor"],
+                    "uncertainty_reduction": row["uncertainty_reduction"],
                     "relativity": row["relativity"],
                 })
         return pl.DataFrame(rows)
@@ -215,22 +225,21 @@ class BayesianRelativities:
         The credible intervals for these segments will be wide.
 
         Practical use: flag thin segments in the rate table for underwriter review.
-        A segment with credibility_factor = 0.1 should not drive a large rate change
-        even if the posterior median relativity differs from 1.0 significantly.
+        A segment with uncertainty_reduction = 0.1 should not drive a large rate
+        change even if the posterior median relativity differs from 1.0 significantly.
 
         Args:
-            credibility_threshold: Segments with credibility_factor below this
-                value are returned. Default 0.3 (meaning less than 30% of the
-                estimate comes from own experience).
+            credibility_threshold: Segments with uncertainty_reduction below this
+                value are returned. Default 0.3.
 
         Returns:
-            Polars DataFrame with factor, level, credibility_factor, and relativity.
+            Polars DataFrame with factor, level, uncertainty_reduction, and relativity.
         """
         creds = self.credibility_factors()
         return (
             creds
-            .filter(creds["credibility_factor"] < credibility_threshold)
-            .sort("credibility_factor")
+            .filter(creds["uncertainty_reduction"] < credibility_threshold)
+            .sort("uncertainty_reduction")
         )
 
     def summary(self) -> "pl.DataFrame":

{bayesian_pricing-0.2.3 → bayesian_pricing-0.2.4}/src/bayesian_pricing/severity.py

@@ -179,6 +179,16 @@ class HierarchicalSeverity:
         _validate_columns_present(df, required_cols)
         _validate_positive(df[severity_col], severity_col)
 
+        if weight_col and weight_col in df.columns:
+            weights_check = df[weight_col].to_numpy(dtype=float)
+            if np.any(weights_check <= 0):
+                n_bad = int(np.sum(weights_check <= 0))
+                raise ValueError(
+                    f"weight_col '{weight_col}' contains {n_bad} zero or negative "
+                    f"value(s). Weights must be strictly positive claim counts. "
+                    f"Remove or exclude zero-claim segments before fitting."
+                )
+
         _check_pymc()
         import pymc as pm
 
@@ -274,9 +284,11 @@ class HierarchicalSeverity:
             mu_severity = pm.math.exp(mu_log)
 
             # Gamma shape: controls within-segment CV.
-            # HalfNormal(2) allows shape from near-0 (high dispersion) to ~6 (low CV).
-            # UK attritional motor claims typically have shape ~0.25-0.7.
-            gamma_shape = pm.HalfNormal("gamma_shape", sigma=2.0)
+            # UK attritional motor severity has CV ~1.5-2.5, implying shape ~0.16-0.44.
+            # HalfNormal(sigma=0.5) has median ~0.34, placing most prior mass in the
+            # attritional motor range. The previous default of sigma=2 put ~83% of
+            # mass above shape=0.44, systematically underestimating dispersion.
+            gamma_shape = pm.HalfNormal("gamma_shape", sigma=0.5)
 
             # Gamma parameterisation: mean = mu, concentration = gamma_shape
             # Each segment observation is a sample average over `weights` claims.

{bayesian_pricing-0.2.3 → bayesian_pricing-0.2.4}/tests/test_relativities.py

@@ -106,7 +106,7 @@ class TestBayesianRelativitiesOutput:
     def test_relativity_table_has_required_columns(self, fitted_rel):
         rt = fitted_rel.relativities(factor="veh_group")
         required_cols = {"level", "relativity", "lower_90pct", "upper_90pct",
-                         "credibility_factor", "interval_width"}
+                         "uncertainty_reduction", "interval_width"}
         assert required_cols.issubset(set(rt.table.columns))
 
     def test_relativity_table_is_polars(self, fitted_rel):
@@ -128,8 +128,8 @@ class TestBayesianRelativitiesOutput:
     def test_credibility_factors_bounded(self, fitted_rel):
         cred = fitted_rel.credibility_factors()
         assert isinstance(cred, pl.DataFrame)
-        assert (cred["credibility_factor"] >= 0).all()
-        assert (cred["credibility_factor"] <= 1).all()
+        assert (cred["uncertainty_reduction"] >= 0).all()
+        assert (cred["uncertainty_reduction"] <= 1).all()
 
     def test_base_level_normalisation(self, freq_segment_data):
         """With base_level={"veh_group": "A"}, group A should have relativity 1.0."""
@@ -178,13 +178,13 @@ class TestBayesianRelativitiesOutput:
         thin = fitted_rel.thin_segments(credibility_threshold=0.5)
         assert isinstance(thin, pl.DataFrame)
         if len(thin) > 0:
-            assert (thin["credibility_factor"] < 0.5).all()
+            assert (thin["uncertainty_reduction"] < 0.5).all()
 
     def test_thin_segments_sorted_ascending(self, fitted_rel):
         thin = fitted_rel.thin_segments(credibility_threshold=1.0)  # all segments
         if len(thin) > 1:
-            cred = thin["credibility_factor"].to_numpy()
-            assert (cred[1:] >= cred[:-1]).all(), "thin_segments should be sorted by credibility"
+            cred = thin["uncertainty_reduction"].to_numpy()
+            assert (cred[1:] >= cred[:-1]).all(), "thin_segments should be sorted by uncertainty_reduction"
 
     def test_single_factor_returns_relativity_table_not_dict(self, fitted_rel):
         result = fitted_rel.relativities(factor="veh_group")

{bayesian_pricing-0.2.3 → bayesian_pricing-0.2.4}/uv.lock

@@ -21,115 +21,37 @@ resolution-markers = [
 name = "arviz"
 version = "0.23.4"
 source = { registry = "https://pypi.org/simple" }
-resolution-markers = [
-    "python_full_version == '3.11.*' and sys_platform == 'win32'",
-    "python_full_version == '3.11.*' and sys_platform == 'emscripten'",
-    "python_full_version == '3.11.*' and sys_platform != 'emscripten' and sys_platform != 'win32'",
-    "python_full_version < '3.11'",
-]
 dependencies = [
-    { name = "h5netcdf", marker = "python_full_version < '3.12'" },
-    { name = "h5py", marker = "python_full_version < '3.12'" },
-    { name = "matplotlib", marker = "python_full_version < '3.12'" },
+    { name = "h5netcdf" },
+    { name = "h5py" },
+    { name = "matplotlib" },
     { name = "numpy", version = "2.2.6", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
-    { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version == '3.11.*'" },
-    { name = "packaging", marker = "python_full_version < '3.12'" },
+    { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },
+    { name = "packaging" },
     { name = "pandas", version = "2.3.3", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
-    { name = "pandas", version = "3.0.1", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version == '3.11.*'" },
-    { name = "platformdirs", marker = "python_full_version < '3.12'" },
+    { name = "pandas", version = "3.0.1", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },
+    { name = "platformdirs" },
     { name = "scipy", version = "1.15.3", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
-    { name = "scipy", version = "1.17.1", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version == '3.11.*'" },
-    { name = "setuptools", marker = "python_full_version < '3.12'" },
-    { name = "typing-extensions", marker = "python_full_version < '3.12'" },
+    { name = "scipy", version = "1.17.1", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },
+    { name = "setuptools" },
+    { name = "typing-extensions" },
     { name = "xarray", version = "2025.6.1", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
-    { name = "xarray", version = "2026.2.0", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version == '3.11.*'" },
+    { name = "xarray", version = "2026.2.0", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },
     { name = "xarray-einstats", version = "0.8.0", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
     { name = "xarray-einstats", version = "0.9.1", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version == '3.11.*'" },
+    { name = "xarray-einstats", version = "0.10.0", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.12'" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/f3/c9/9c853633715f972eecc20995763c6e3005a3afcdcf47e39d20cd1c2889cd/arviz-0.23.4.tar.gz", hash = "sha256:611be826995066036c9443ea98d11486c279ef3da3b6cdc5c0816fab434115b9", size = 1592968, upload-time = "2026-02-04T17:57:53.664Z" }
 wheels = [
     { url = "https://files.pythonhosted.org/packages/44/1f/227f9cb7edcd3e14ab05928f3db00e9d595c0f269c87bf35f565ce44941b/arviz-0.23.4-py3-none-any.whl", hash = "sha256:c46c7faf8a06abadc9b5b64000584062ecbc20c2298e2bd6dfba04bb01a684ca", size = 1673773, upload-time = "2026-02-04T17:57:51.778Z" },
 ]
 
-[[package]]
-name = "arviz"
-version = "1.0.0"
-source = { registry = "https://pypi.org/simple" }
-resolution-markers = [
-    "python_full_version >= '3.14' and sys_platform == 'win32'",
-    "python_full_version >= '3.14' and sys_platform == 'emscripten'",
-    "python_full_version >= '3.14' and sys_platform != 'emscripten' and sys_platform != 'win32'",
-    "python_full_version == '3.13.*' and sys_platform == 'win32'",
-    "python_full_version == '3.12.*' and sys_platform == 'win32'",
-    "python_full_version == '3.13.*' and sys_platform == 'emscripten'",
-    "python_full_version == '3.12.*' and sys_platform == 'emscripten'",
-    "python_full_version == '3.13.*' and sys_platform != 'emscripten' and sys_platform != 'win32'",
-    "python_full_version == '3.12.*' and sys_platform != 'emscripten' and sys_platform != 'win32'",
-]
-dependencies = [
-    { name = "arviz-base", marker = "python_full_version >= '3.12'" },
-    { name = "arviz-plots", marker = "python_full_version >= '3.12'" },
-    { name = "arviz-stats", extra = ["xarray"], marker = "python_full_version >= '3.12'" },
-]
-sdist = { url = "https://files.pythonhosted.org/packages/5b/e1/398fcbc5b043245fe7d704346b42385b886956a99fcb09964b41aada4267/arviz-1.0.0.tar.gz", hash = "sha256:9bc978e8b590d5ee88fe024c4fa10e214dc42c31329e645b2ad998419cbe7ff6", size = 8272, upload-time = "2026-03-02T14:58:48.996Z" }
-wheels = [
-    { url = "https://files.pythonhosted.org/packages/30/dc/b1dcb62745f58f800d4073c9ad914a5cf882fe1d8f20793eae0a1d54bbac/arviz-1.0.0-py3-none-any.whl", hash = "sha256:8caf30516bb7e13c237d266e3d37c59531898a93db9e89c1f1bb9320699e97dd", size = 8445, upload-time = "2026-03-02T14:58:47.7Z" },
-]
-
-[[package]]
-name = "arviz-base"
-version = "1.0.0"
-source = { registry = "https://pypi.org/simple" }
-dependencies = [
-    { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.12'" },
-    { name = "typing-extensions", marker = "python_full_version >= '3.12'" },
-    { name = "xarray", version = "2026.2.0", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.12'" },
-]
-sdist = { url = "https://files.pythonhosted.org/packages/17/9b/84b06b529b1b397c50f8ff2e5af5af7a09eef24c6ee5334e08e2cf12ef96/arviz_base-1.0.0.tar.gz", hash = "sha256:6b9796043b4e394056fd8a1ab51c8d4d3e88dcf8f2b698a46b660ff4346c8841", size = 1403207, upload-time = "2026-03-02T07:36:17.604Z" }
-wheels = [
-    { url = "https://files.pythonhosted.org/packages/35/69/6624d40ec780a30d187e9224e1c2bf51eebe0e61ebd2c7dee06dd4464a24/arviz_base-1.0.0-py3-none-any.whl", hash = "sha256:77c93e3503166517da8155d580ec9615e9b7b89ab7890f8d9ab72c78a9b6c9e6", size = 1420656, upload-time = "2026-03-02T07:36:15.588Z" },
-]
-
-[[package]]
-name = "arviz-plots"
-version = "1.0.0"
-source = { registry = "https://pypi.org/simple" }
-dependencies = [
-    { name = "arviz-base", marker = "python_full_version >= '3.12'" },
-    { name = "arviz-stats", extra = ["xarray"], marker = "python_full_version >= '3.12'" },
-]
-sdist = { url = "https://files.pythonhosted.org/packages/12/af/113e16c1c7990badc433d694008ffb6d235265d9da15c22bff1ca485d4b3/arviz_plots-1.0.0.tar.gz", hash = "sha256:e8bb70824b50f09f7c527dac4519aef1b72cb0a9eadcd4da25b80c22a4bcef71", size = 143729, upload-time = "2026-03-02T10:03:36.015Z" }
-wheels = [
-    { url = "https://files.pythonhosted.org/packages/aa/ff/9e1914853bd6df5f7ddf7a1cc261baeae87992bd46c9b40d0300b499d18d/arviz_plots-1.0.0-py3-none-any.whl", hash = "sha256:b086a603be6d880f6b446beae4d2111c7c3dc2ce2e65481cd88097f474694dde", size = 225351, upload-time = "2026-03-02T10:03:34.437Z" },
-]
-
-[[package]]
-name = "arviz-stats"
-version = "1.0.0"
-source = { registry = "https://pypi.org/simple" }
-dependencies = [
-    { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.12'" },
-    { name = "scipy", version = "1.17.1", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.12'" },
-]
-sdist = { url = "https://files.pythonhosted.org/packages/5b/09/df9f50fd79ca7a990564bfb45902ad79b5d87ed7deb6ae0f90d3844c5408/arviz_stats-1.0.0.tar.gz", hash = "sha256:056228574bbc1e9045fd41da168d62fe4167097fd886056b8085a40799142e53", size = 144731, upload-time = "2026-03-02T08:06:52.294Z" }
-wheels = [
-    { url = "https://files.pythonhosted.org/packages/b2/3f/7f0df19fc24f375b2e28e594893a1869314b868a6791046f1a56b5ca78a1/arviz_stats-1.0.0-py3-none-any.whl", hash = "sha256:3b2e5e8c8714827497e82eb1ab8064a16a7fb1a8f733edf325bff2c81ec4a400", size = 171588, upload-time = "2026-03-02T08:06:50.685Z" },
-]
-
-[package.optional-dependencies]
-xarray = [
-    { name = "arviz-base", marker = "python_full_version >= '3.12'" },
-    { name = "xarray", version = "2026.2.0", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.12'" },
-    { name = "xarray-einstats", version = "0.10.0", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.12'" },
-]
-
 [[package]]
 name = "bayesian-pricing"
-version = "0.2.0"
+version = "0.2.1"
 source = { editable = "." }
 dependencies = [
-    { name = "arviz", version = "0.23.4", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.12'" },
-    { name = "arviz", version = "1.0.0", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.12'" },
+    { name = "arviz" },
     { name = "numpy", version = "2.2.6", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
     { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },
     { name = "pandas", version = "2.3.3", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
@@ -161,12 +83,12 @@ dev = [
 
 [package.metadata]
 requires-dist = [
-    { name = "arviz", specifier = ">=0.17" },
+    { name = "arviz", specifier = ">=0.17,<1.0" },
     { name = "jax", marker = "extra == 'numpyro'", specifier = ">=0.4" },
     { name = "numpy", specifier = ">=1.24" },
     { name = "numpyro", marker = "extra == 'numpyro'", specifier = ">=0.13" },
     { name = "pandas", specifier = ">=2.0" },
-    { name = "polars", specifier = ">=0.20" },
+    { name = "polars", specifier = ">=1.0" },
     { name = "pyarrow", specifier = ">=23.0.1" },
     { name = "pymc", marker = "extra == 'pymc'", specifier = ">=5.0" },
     { name = "pytest", marker = "extra == 'dev'", specifier = ">=7.0" },
@@ -291,12 +213,21 @@ name = "contourpy"
 version = "1.3.3"
 source = { registry = "https://pypi.org/simple" }
 resolution-markers = [
+    "python_full_version >= '3.14' and sys_platform == 'win32'",
+    "python_full_version >= '3.14' and sys_platform == 'emscripten'",
+    "python_full_version >= '3.14' and sys_platform != 'emscripten' and sys_platform != 'win32'",
+    "python_full_version == '3.13.*' and sys_platform == 'win32'",
+    "python_full_version == '3.12.*' and sys_platform == 'win32'",
+    "python_full_version == '3.13.*' and sys_platform == 'emscripten'",
+    "python_full_version == '3.12.*' and sys_platform == 'emscripten'",
+    "python_full_version == '3.13.*' and sys_platform != 'emscripten' and sys_platform != 'win32'",
+    "python_full_version == '3.12.*' and sys_platform != 'emscripten' and sys_platform != 'win32'",
     "python_full_version == '3.11.*' and sys_platform == 'win32'",
     "python_full_version == '3.11.*' and sys_platform == 'emscripten'",
     "python_full_version == '3.11.*' and sys_platform != 'emscripten' and sys_platform != 'win32'",
 ]
 dependencies = [
-    { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version == '3.11.*'" },
+    { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/58/01/1253e6698a07380cd31a736d248a3f2a50a7c88779a1813da27503cadc2a/contourpy-1.3.3.tar.gz", hash = "sha256:083e12155b210502d0bca491432bb04d56dc3432f95a979b429f2848c3dbe880", size = 13466174, upload-time = "2025-07-26T12:03:12.549Z" }
 wheels = [
@@ -597,8 +528,8 @@ version = "1.8.1"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "numpy", version = "2.2.6", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
-    { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version == '3.11.*'" },
-    { name = "packaging", marker = "python_full_version < '3.12'" },
+    { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },
+    { name = "packaging" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/ef/03/92d6cc02c0055158167255980461155d6e17f1c4143c03f8bcc18d3e3f3a/h5netcdf-1.8.1.tar.gz", hash = "sha256:9b396a4cc346050fc1a4df8523bc1853681ec3544e0449027ae397cb953c7a16", size = 78679, upload-time = "2026-01-23T07:35:31.233Z" }
 wheels = [
@@ -611,7 +542,7 @@ version = "3.16.0"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "numpy", version = "2.2.6", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
-    { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version == '3.11.*'" },
+    { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/db/33/acd0ce6863b6c0d7735007df01815403f5589a21ff8c2e1ee2587a38f548/h5py-3.16.0.tar.gz", hash = "sha256:a0dbaad796840ccaa67a4c144a0d0c8080073c34c76d5a6941d6818678ef2738", size = 446526, upload-time = "2026-03-06T13:49:08.07Z" }
 wheels = [
@@ -970,16 +901,16 @@ version = "3.10.8"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "contourpy", version = "1.3.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
-    { name = "contourpy", version = "1.3.3", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version == '3.11.*'" },
-    { name = "cycler", marker = "python_full_version < '3.12'" },
-    { name = "fonttools", marker = "python_full_version < '3.12'" },
-    { name = "kiwisolver", marker = "python_full_version < '3.12'" },
+    { name = "contourpy", version = "1.3.3", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },
+    { name = "cycler" },
+    { name = "fonttools" },
+    { name = "kiwisolver" },
     { name = "numpy", version = "2.2.6", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
-    { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version == '3.11.*'" },
-    { name = "packaging", marker = "python_full_version < '3.12'" },
-    { name = "pillow", marker = "python_full_version < '3.12'" },
-    { name = "pyparsing", marker = "python_full_version < '3.12'" },
-    { name = "python-dateutil", marker = "python_full_version < '3.12'" },
+    { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },
+    { name = "packaging" },
+    { name = "pillow" },
+    { name = "pyparsing" },
+    { name = "python-dateutil" },
 ]
 sdist = { url = "https://files.pythonhosted.org/packages/8a/76/d3c6e3a13fe484ebe7718d14e269c9569c4eb0020a968a327acb3b9a8fe6/matplotlib-3.10.8.tar.gz", hash = "sha256:2299372c19d56bcd35cf05a2738308758d32b9eaed2371898d8f5bd33f084aa3", size = 34806269, upload-time = "2025-12-10T22:56:51.155Z" }
 wheels = [
@@ -1733,7 +1664,7 @@ resolution-markers = [
     "python_full_version < '3.11'",
 ]
 dependencies = [
-    { name = "arviz", version = "0.23.4", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
+    { name = "arviz", marker = "python_full_version < '3.11'" },
     { name = "cachetools", marker = "python_full_version < '3.11'" },
     { name = "cloudpickle", marker = "python_full_version < '3.11'" },
     { name = "numpy", version = "2.2.6", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version < '3.11'" },
@@ -1768,8 +1699,7 @@ resolution-markers = [
     "python_full_version == '3.11.*' and sys_platform != 'emscripten' and sys_platform != 'win32'",
 ]
 dependencies = [
-    { name = "arviz", version = "0.23.4", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version == '3.11.*'" },
-    { name = "arviz", version = "1.0.0", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.12'" },
+    { name = "arviz", marker = "python_full_version >= '3.11'" },
     { name = "cachetools", marker = "python_full_version >= '3.11'" },
     { name = "cloudpickle", marker = "python_full_version >= '3.11'" },
     { name = "numpy", version = "2.4.2", source = { registry = "https://pypi.org/simple" }, marker = "python_full_version >= '3.11'" },