bossanova 0.1.0.dev9__tar.gz → 0.1.0.dev11__tar.gz

{bossanova-0.1.0.dev9 → bossanova-0.1.0.dev11}/.gitignore

@@ -81,3 +81,4 @@ bossanova-docs/help/
 bossanova-docs/wip/
 .hypothesis/
 research.md
+201b-ghct-05-models/

{bossanova-0.1.0.dev9 → bossanova-0.1.0.dev11}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: bossanova
-Version: 0.1.0.dev9
+Version: 0.1.0.dev11
 Summary: Bridging statistical cultures with some jazz
 Author: Eshin Jolly
 License-Expression: MIT
@@ -9,6 +9,8 @@ Keywords: analysis,multi-level-modeling,regression,statistics
 Classifier: Intended Audience :: Science/Research
 Classifier: Operating System :: OS Independent
 Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
 Requires-Python: >=3.11
 Requires-Dist: attrs>=24.2.0
 Requires-Dist: jax>=0.8.0
@@ -28,20 +30,16 @@ Description-Content-Type: text/markdown
 
 Unified mixed-effects modeling in Python. JAX + NumPy backends. R/lme4 parity.
 
-## Quick Example
+## Installation
 
-```python
-from bossanova import model, load_dataset
+```bash
+uv add bossanova
+```
 
-df = load_dataset("sleepstudy")
+or
 
-# Fit a mixed model
-m = model("Reaction ~ Days + (1 + Days | Subject)", data=df).fit().infer()
-m.params
-
-# Explore marginal effects
-m.explore("Days").infer()
-m.effects
+```bash
+pip install bossanova
 ```
 
 ## Features
@@ -50,15 +48,26 @@ m.effects
 - **R-style formulas** — `y ~ x + (1|group)`, `factor()`, `center()`, `scale()`
 - **Multiple inference strategies** — asymptotic, bootstrap, permutation, cross-validation
 - **Marginal effects** — `.explore()` for estimated marginal means, slopes, and contrasts
-- **JAX acceleration** — automatic GPU/TPU acceleration when JAX is available, NumPy fallback
+- **JAX acceleration** — automatic acceleration when JAX is available, NumPy fallback
 - **R parity** — matches lme4/emmeans within f64 tolerance
 
-## Installation
+## Quick Example
 
-```bash
-pip install bossanova
+```python
+from bossanova import model, load_dataset
+
+df = load_dataset("sleepstudy")
+
+# Fit a mixed model
+m = model("Reaction ~ Days + (1 + Days | Subject)", data=df).fit()
+m.params
+
+# Perform bootstrapped parameter inference
+m.infer(how='boot')
+
+# Explore marginal effects
+m.explore("Days")
+m.effects
 ```
 
-## Documentation
 
-See the [full documentation](https://bossanova.readthedocs.io/) for tutorials, API reference, and examples.

{bossanova-0.1.0.dev9 → bossanova-0.1.0.dev11}/README.md

@@ -2,20 +2,16 @@
 
 Unified mixed-effects modeling in Python. JAX + NumPy backends. R/lme4 parity.
 
-## Quick Example
+## Installation
 
-```python
-from bossanova import model, load_dataset
+```bash
+uv add bossanova
+```
 
-df = load_dataset("sleepstudy")
+or
 
-# Fit a mixed model
-m = model("Reaction ~ Days + (1 + Days | Subject)", data=df).fit().infer()
-m.params
-
-# Explore marginal effects
-m.explore("Days").infer()
-m.effects
+```bash
+pip install bossanova
 ```
 
 ## Features
@@ -24,15 +20,26 @@ m.effects
 - **R-style formulas** — `y ~ x + (1|group)`, `factor()`, `center()`, `scale()`
 - **Multiple inference strategies** — asymptotic, bootstrap, permutation, cross-validation
 - **Marginal effects** — `.explore()` for estimated marginal means, slopes, and contrasts
-- **JAX acceleration** — automatic GPU/TPU acceleration when JAX is available, NumPy fallback
+- **JAX acceleration** — automatic acceleration when JAX is available, NumPy fallback
 - **R parity** — matches lme4/emmeans within f64 tolerance
 
-## Installation
+## Quick Example
 
-```bash
-pip install bossanova
+```python
+from bossanova import model, load_dataset
+
+df = load_dataset("sleepstudy")
+
+# Fit a mixed model
+m = model("Reaction ~ Days + (1 + Days | Subject)", data=df).fit()
+m.params
+
+# Perform bootstrapped parameter inference
+m.infer(how='boot')
+
+# Explore marginal effects
+m.explore("Days")
+m.effects
 ```
 
-## Documentation
 
-See the [full documentation](https://bossanova.readthedocs.io/) for tutorials, API reference, and examples.

{bossanova-0.1.0.dev9 → bossanova-0.1.0.dev11}/bossanova/internal/containers/builders/dataframes.py

@@ -15,6 +15,13 @@ import polars as pl
 from bossanova.internal.containers.builders.results import (
     append_inference_columns,
 )
+from bossanova.internal.containers.schemas import (
+    Col,
+    ParamsAsymp,
+    ParamsBase,
+    ParamsCv,
+    VaryingCorrSchema,
+)
 
 if TYPE_CHECKING:
     from bossanova.internal.containers.structs import (
@@ -60,21 +67,22 @@ def build_params_dataframe(
     terms = list(bundle.X_names)
     estimates = fit.coef.tolist()
 
-    data: dict[str, list] = {"term": terms, "estimate": estimates}
+    data: dict[str, list] = {Col.TERM: terms, Col.ESTIMATE: estimates}
 
     if params_inference is not None:
         if params_inference.method == "cv":
-            data["pre"] = params_inference.pre.tolist()
-            data["pre_sd"] = params_inference.pre_sd.tolist()
-        else:
-            data["se"] = params_inference.se.tolist()
-            data["statistic"] = params_inference.statistic.tolist()
-            data["df"] = params_inference.df.tolist()
-            data["p_value"] = params_inference.p_value.tolist()
-            data["ci_lower"] = params_inference.ci_lower.tolist()
-            data["ci_upper"] = params_inference.ci_upper.tolist()
+            data[Col.PRE] = params_inference.pre.tolist()
+            data[Col.PRE_SD] = params_inference.pre_sd.tolist()
+            return pl.DataFrame(data, schema=ParamsCv)
+        data[Col.SE] = params_inference.se.tolist()
+        data[Col.STATISTIC] = params_inference.statistic.tolist()
+        data[Col.DF] = params_inference.df.tolist()
+        data[Col.P_VALUE] = params_inference.p_value.tolist()
+        data[Col.CI_LOWER] = params_inference.ci_lower.tolist()
+        data[Col.CI_UPPER] = params_inference.ci_upper.tolist()
+        return pl.DataFrame(data, schema=ParamsAsymp)
 
-    return pl.DataFrame(data)
+    return pl.DataFrame(data, schema=ParamsBase)
 
 
 def build_varying_offsets_dataframe(
@@ -99,11 +107,11 @@ def build_varying_offsets_dataframe(
             if effect_name in varying_offsets.pi_lower:
                 result = result.with_columns(
                     pl.Series(
-                        f"pi_lower_{effect_name}",
+                        f"{Col.PI_LOWER_PREFIX}{effect_name}",
                         varying_offsets.pi_lower[effect_name],
                     ),
                     pl.Series(
-                        f"pi_upper_{effect_name}",
+                        f"{Col.PI_UPPER_PREFIX}{effect_name}",
                         varying_offsets.pi_upper[effect_name],
                     ),
                 )
@@ -159,15 +167,15 @@ def build_varying_spread_dataframe(
         ci_upper_vals = []
         ci_method_vals = []
 
-        for comp_name in result["component"].to_list():
+        for comp_name in result[Col.COMPONENT].to_list():
             ci_lower_vals.append(varying_spread.ci_lower.get(comp_name, float("nan")))
             ci_upper_vals.append(varying_spread.ci_upper.get(comp_name, float("nan")))
             ci_method_vals.append(varying_spread.ci_method or "")
 
         result = result.with_columns(
-            pl.Series("ci_lower", ci_lower_vals),
-            pl.Series("ci_upper", ci_upper_vals),
-            pl.Series("ci_method", ci_method_vals),
+            pl.Series(Col.CI_LOWER, ci_lower_vals),
+            pl.Series(Col.CI_UPPER, ci_upper_vals),
+            pl.Series(Col.CI_METHOD, ci_method_vals),
         )
 
     return result
@@ -214,11 +222,12 @@ def build_varying_corr_dataframe(
 
     return pl.DataFrame(
         {
-            "group": groups,
-            "effect1": effect1s,
-            "effect2": effect2s,
-            "corr": corrs,
-        }
+            Col.GROUP: groups,
+            Col.EFFECT1: effect1s,
+            Col.EFFECT2: effect2s,
+            Col.CORR: corrs,
+        },
+        schema=VaryingCorrSchema,
     )
 
 
@@ -248,7 +257,7 @@ def build_effects_dataframe(mee: MeeState) -> pl.DataFrame:
         columns.
     """
     result = mee.grid.clone()
-    result = result.with_columns(pl.Series("estimate", mee.estimate))
+    result = result.with_columns(pl.Series(Col.ESTIMATE, mee.estimate))
     return append_inference_columns(result, mee)
 
 
@@ -262,15 +271,15 @@ def build_joint_test_dataframe(state: JointTestState) -> pl.DataFrame:
         DataFrame with ``term``, ``df1``, optional ``df2``,
         ``f_ratio`` or ``Chisq``, and ``p_value`` columns.
     """
-    result = pl.DataFrame({"term": list(state.terms)})
-    result = result.with_columns(pl.Series("df1", state.df1))
+    result = pl.DataFrame({Col.TERM: list(state.terms)})
+    result = result.with_columns(pl.Series(Col.DF1, state.df1))
 
     if state.df2 is not None:
-        result = result.with_columns(pl.Series("df2", state.df2))
+        result = result.with_columns(pl.Series(Col.DF2, state.df2))
 
-    stat_name = "f_ratio" if state.test_type == "F" else "Chisq"
+    stat_name = Col.F_RATIO if state.test_type == "F" else Col.CHISQ
     result = result.with_columns(pl.Series(stat_name, state.statistic))
-    result = result.with_columns(pl.Series("p_value", state.p_value))
+    result = result.with_columns(pl.Series(Col.P_VALUE, state.p_value))
 
     return result
 
@@ -286,18 +295,18 @@ def build_predictions_dataframe(pred: PredictionState) -> pl.DataFrame:
         DataFrame with ``fitted``, optional ``link``, inference columns,
         and optional CV columns.
     """
-    data: dict[str, list] = {"fitted": pred.fitted.tolist()}
+    data: dict[str, list] = {Col.FITTED: pred.fitted.tolist()}
     if pred.link is not None:
-        data["link"] = pred.link.tolist()
+        data[Col.LINK] = pred.link.tolist()
     result = pl.DataFrame(data)
 
     result = append_inference_columns(result, pred)
 
     if pred.has_cv:
         result = result.with_columns(
-            pl.Series("cv_fitted", pred.cv_fitted),
-            pl.Series("cv_residual", pred.cv_residual),
-            pl.Series("cv_fold", pred.cv_fold),
+            pl.Series(Col.CV_FITTED, pred.cv_fitted),
+            pl.Series(Col.CV_RESIDUAL, pred.cv_residual),
+            pl.Series(Col.CV_FOLD, pred.cv_fold),
         )
 
     return result
@@ -321,9 +330,11 @@ def build_simulations_dataframe(
 
     if sim_inference is not None:
         if sim_inference.sim_mean is not None:
-            result = result.with_columns(pl.Series("sim_mean", sim_inference.sim_mean))
+            result = result.with_columns(
+                pl.Series(Col.SIM_MEAN, sim_inference.sim_mean)
+            )
         if sim_inference.sim_sd is not None:
-            result = result.with_columns(pl.Series("sim_sd", sim_inference.sim_sd))
+            result = result.with_columns(pl.Series(Col.SIM_SD, sim_inference.sim_sd))
         for q_name, q_values in sim_inference.sim_quantiles.items():
             result = result.with_columns(pl.Series(q_name, q_values))
 

{bossanova-0.1.0.dev9 → bossanova-0.1.0.dev11}/bossanova/internal/containers/builders/results.py

@@ -10,18 +10,20 @@ from __future__ import annotations
 
 import polars as pl
 
+from bossanova.internal.containers.schemas import Col
+
 __all__ = [
     "append_inference_columns",
 ]
 
 # Standard inference column names, in canonical display order.
 _INFERENCE_COLS: tuple[str, ...] = (
-    "se",
-    "statistic",
-    "df",
-    "p_value",
-    "ci_lower",
-    "ci_upper",
+    Col.SE,
+    Col.STATISTIC,
+    Col.DF,
+    Col.P_VALUE,
+    Col.CI_LOWER,
+    Col.CI_UPPER,
 )
 
 

{bossanova-0.1.0.dev9 → bossanova-0.1.0.dev11}/bossanova/internal/containers/builders/state.py

@@ -6,6 +6,11 @@ from typing import TYPE_CHECKING
 
 import numpy as np
 
+from bossanova.internal.containers.schemas import (
+    Col,
+    ResamplesBootSchema,
+    ResamplesPermSchema,
+)
 from bossanova.internal.containers.structs.state import (
     CVState,
     FitState,
@@ -708,16 +713,17 @@ def build_resamples_dataframe(rs: ResamplesState) -> pl.DataFrame:
 
         return pl.DataFrame(
             {
-                "term": names,
-                "observed": observed.tolist(),
-                "mean": means.tolist(),
-                "se": ses.tolist(),
-                "ci_lower": ci_lower.tolist(),
-                "ci_upper": ci_upper.tolist(),
-                "n_resamples": [rs.n_resamples] * len(names),
-                "method": [rs.method] * len(names),
-                "context": [rs.context] * len(names),
-            }
+                Col.TERM: names,
+                Col.OBSERVED: observed.tolist(),
+                Col.MEAN: means.tolist(),
+                Col.SE: ses.tolist(),
+                Col.CI_LOWER: ci_lower.tolist(),
+                Col.CI_UPPER: ci_upper.tolist(),
+                Col.N_RESAMPLES: [rs.n_resamples] * len(names),
+                Col.METHOD: [rs.method] * len(names),
+                Col.CONTEXT: [rs.context] * len(names),
+            },
+            schema=ResamplesBootSchema,
         )
 
     # Permutation: compute two-sided p-values
@@ -725,11 +731,12 @@ def build_resamples_dataframe(rs: ResamplesState) -> pl.DataFrame:
 
     return pl.DataFrame(
         {
-            "term": names,
-            "observed": observed.tolist(),
-            "p_value": p_values.tolist(),
-            "n_resamples": [rs.n_resamples] * len(names),
-            "method": [rs.method] * len(names),
-            "context": [rs.context] * len(names),
-        }
+            Col.TERM: names,
+            Col.OBSERVED: observed.tolist(),
+            Col.P_VALUE: p_values.tolist(),
+            Col.N_RESAMPLES: [rs.n_resamples] * len(names),
+            Col.METHOD: [rs.method] * len(names),
+            Col.CONTEXT: [rs.context] * len(names),
+        },
+        schema=ResamplesPermSchema,
     )