combatlearn 0.1.1__tar.gz → 0.1.2__tar.gz

{combatlearn-0.1.1/src/combatlearn.egg-info → combatlearn-0.1.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: combatlearn
-Version: 0.1.1
+Version: 0.1.2
 Summary: Batch-effect harmonization for machine learning frameworks.
 Author-email: Ettore Rocchi <ettoreroc@gmail.com>
 License: MIT License
@@ -42,10 +42,10 @@ Dynamic: license-file
 
 # **combatlearn**
 
-[![Python versions](https://img.shields.io/badge/python-3.10%20%7C%203.11%20%7C%203.12-blue?logo=python)](https://img.shields.io/badge/python-3.10%20%7C%203.11%20%7C%203.12-blue?logo=python)
+[![Python versions](https://img.shields.io/badge/python-%3E%3D3.10-blue?logo=python)](https://www.python.org/)
 [![Test](https://github.com/EttoreRocchi/combatlearn/actions/workflows/test.yaml/badge.svg)](https://github.com/EttoreRocchi/combatlearn/actions/workflows/test.yaml)
 [![PyPI Downloads](https://static.pepy.tech/badge/combatlearn)](https://pepy.tech/projects/combatlearn)
-[![PyPI version](https://badge.fury.io/py/combatlearn.svg)](https://pypi.org/project/combatlearn/)
+[![PyPI Version](https://img.shields.io/pypi/v/combatlearn?cacheSeconds=300)](https://pypi.org/project/combatlearn/)
 [![License](https://img.shields.io/github/license/EttoreRocchi/combatlearn)](https://github.com/EttoreRocchi/combatlearn/blob/main/LICENSE)
 
 <div align="center">
@@ -56,7 +56,7 @@ Dynamic: license-file
 
 **Three methods**:
 - `method="johnson"` - classic ComBat (Johnson _et al._, 2007)
-- `method="fortin"` - covariate-aware ComBat (Fortin _et al._, 2018)
+- `method="fortin"` - neuroComBat (Fortin _et al._, 2018)
 - `method="chen"` - CovBat (Chen _et al._, 2022)
 
 ## Installation

{combatlearn-0.1.1 → combatlearn-0.1.2}/README.md

@@ -1,9 +1,9 @@
 # **combatlearn**
 
-[![Python versions](https://img.shields.io/badge/python-3.10%20%7C%203.11%20%7C%203.12-blue?logo=python)](https://img.shields.io/badge/python-3.10%20%7C%203.11%20%7C%203.12-blue?logo=python)
+[![Python versions](https://img.shields.io/badge/python-%3E%3D3.10-blue?logo=python)](https://www.python.org/)
 [![Test](https://github.com/EttoreRocchi/combatlearn/actions/workflows/test.yaml/badge.svg)](https://github.com/EttoreRocchi/combatlearn/actions/workflows/test.yaml)
 [![PyPI Downloads](https://static.pepy.tech/badge/combatlearn)](https://pepy.tech/projects/combatlearn)
-[![PyPI version](https://badge.fury.io/py/combatlearn.svg)](https://pypi.org/project/combatlearn/)
+[![PyPI Version](https://img.shields.io/pypi/v/combatlearn?cacheSeconds=300)](https://pypi.org/project/combatlearn/)
 [![License](https://img.shields.io/github/license/EttoreRocchi/combatlearn)](https://github.com/EttoreRocchi/combatlearn/blob/main/LICENSE)
 
 <div align="center">
@@ -14,7 +14,7 @@
 
 **Three methods**:
 - `method="johnson"` - classic ComBat (Johnson _et al._, 2007)
-- `method="fortin"` - covariate-aware ComBat (Fortin _et al._, 2018)
+- `method="fortin"` - neuroComBat (Fortin _et al._, 2018)
 - `method="chen"` - CovBat (Chen _et al._, 2022)
 
 ## Installation

{combatlearn-0.1.1/src → combatlearn-0.1.2}/combatlearn/combat.py

@@ -139,7 +139,7 @@ class ComBatModel:
 
         if self.reference_batch is not None and self.reference_batch not in batch.cat.categories:
             raise ValueError(
-                f"reference_batch={self.reference_batch!r} not present in the data batches "
+                f"reference_batch={self.reference_batch!r} not present in the data batches."
                 f"{list(batch.cat.categories)}"
             )
 
@@ -218,69 +218,94 @@ class ComBatModel:
         disc: Optional[pd.DataFrame],
         cont: Optional[pd.DataFrame],
     ) -> None:
-        """Fortin et al. (2018) ComBat."""
-        batch_levels = batch.cat.categories
-        n_batch = len(batch_levels)
+        """Fortin et al. (2018) neuroComBat."""
+        self._batch_levels = batch.cat.categories
+        n_batch = len(self._batch_levels)
         n_samples = len(X)
 
-        batch_dummies = pd.get_dummies(batch, drop_first=False)
+        batch_dummies = pd.get_dummies(batch, drop_first=False).astype(float)
+        if self.reference_batch is not None:
+            if self.reference_batch not in self._batch_levels:
+                raise ValueError(
+                    f"reference_batch={self.reference_batch!r} not present in batches."
+                    f"{list(self._batch_levels)}"
+                )
+            batch_dummies.loc[:, self.reference_batch] = 1.0
+
         parts: list[pd.DataFrame] = [batch_dummies]
         if disc is not None:
-            parts.append(pd.get_dummies(disc.astype("category"), drop_first=True))
+            parts.append(
+                pd.get_dummies(
+                    disc.astype("category"), drop_first=True
+                ).astype(float)
+            )
+
         if cont is not None:
-            parts.append(cont)
-        design = pd.concat(parts, axis=1).astype(float).values
+            parts.append(cont.astype(float))
+
+        design = pd.concat(parts, axis=1).values
         p_design = design.shape[1]
 
         X_np = X.values
         beta_hat = la.lstsq(design, X_np, rcond=None)[0]
 
-        gamma_hat = beta_hat[:n_batch]
+        beta_hat_batch = beta_hat[:n_batch]
         self._beta_hat_nonbatch = beta_hat[n_batch:]
 
-        n_per_batch_arr = batch.value_counts().sort_index().values
-        self._n_per_batch = dict(zip(batch_levels, n_per_batch_arr))
+        n_per_batch = batch.value_counts().sort_index().astype(int).values
+        self._n_per_batch = dict(zip(self._batch_levels, n_per_batch))
+
+        if self.reference_batch is not None:
+            ref_idx = list(self._batch_levels).index(self.reference_batch)
+            grand_mean = beta_hat_batch[ref_idx]
+        else:
+            grand_mean = (n_per_batch / n_samples) @ beta_hat_batch
+            ref_idx = None
 
-        grand_mean = (n_per_batch_arr / n_samples) @ gamma_hat
         self._grand_mean = pd.Series(grand_mean, index=X.columns)
 
-        resid = X_np - design @ beta_hat
-        var_pooled = (resid ** 2).sum(axis=0) / (n_samples - p_design) + self.eps
+        if self.reference_batch is not None:
+            ref_mask = (batch == self.reference_batch).values
+            resid = X_np[ref_mask] - design[ref_mask] @ beta_hat
+            denom = int(ref_mask.sum())
+        else:
+            resid = X_np - design @ beta_hat
+            denom = n_samples
+        var_pooled = (resid ** 2).sum(axis=0) / denom + self.eps
         self._pooled_var = pd.Series(var_pooled, index=X.columns)
 
         stand_mean = grand_mean + design[:, n_batch:] @ self._beta_hat_nonbatch
         Xs = (X_np - stand_mean) / np.sqrt(var_pooled)
 
-        delta_hat = np.empty_like(gamma_hat)
-        for i, lvl in enumerate(batch_levels):
-            idx = batch == lvl
-            delta_hat[i] = Xs[idx].var(axis=0, ddof=1) + self.eps
+        gamma_hat = np.vstack(
+            [Xs[batch == lvl].mean(axis=0) for lvl in self._batch_levels]
+        )
+        delta_hat = np.vstack(
+            [Xs[batch == lvl].var(axis=0, ddof=1) + self.eps
+             for lvl in self._batch_levels]
+        )
 
         if self.mean_only:
             gamma_star = self._shrink_gamma(
-                gamma_hat, delta_hat, n_per_batch_arr, parametric=self.parametric
+                gamma_hat, delta_hat, n_per_batch,
+                parametric = self.parametric
             )
             delta_star = np.ones_like(delta_hat)
         else:
             gamma_star, delta_star = self._shrink_gamma_delta(
-                gamma_hat, delta_hat, n_per_batch_arr, parametric=self.parametric
+                gamma_hat, delta_hat, n_per_batch,
+                parametric = self.parametric
             )
 
-        if self.reference_batch is not None:
-            ref_idx = list(batch_levels).index(self.reference_batch)
-            gamma_ref = gamma_star[ref_idx]
-            delta_ref = delta_star[ref_idx]
-            gamma_star = gamma_star - gamma_ref
+        if ref_idx is not None:
+            gamma_star[ref_idx] = 0.0
             if not self.mean_only:
-                delta_star = delta_star / delta_ref
-            self._reference_batch_idx = ref_idx
-        else:
-            self._reference_batch_idx = None
+                delta_star[ref_idx] = 1.0
+        self._reference_batch_idx = ref_idx
 
-        self._batch_levels = batch_levels
         self._gamma_star = gamma_star
         self._delta_star = delta_star
-        self._n_batch = n_batch
+        self._n_batch  = n_batch
         self._p_design = p_design
     
     def _fit_chen(
@@ -434,7 +459,7 @@ class ComBatModel:
         *,
         parametric: bool,
     ) -> FloatArray:
-        """Convenience wrapper that returns only γ⋆ (for *mean‑only* mode)."""
+        """Convenience wrapper that returns only γ⋆ (for *mean-only* mode)."""
         gamma, _ = self._shrink_gamma_delta(gamma_hat, delta_hat, n_per_batch, parametric=parametric)
         return gamma
 
@@ -454,7 +479,7 @@ class ComBatModel:
         batch = self._as_series(batch, idx, "batch")
         unseen = set(batch.cat.categories) - set(self._batch_levels)
         if unseen:
-            raise ValueError(f"Unseen batch levels during transform: {unseen}")
+            raise ValueError(f"Unseen batch levels during transform: {unseen}.")
         disc = self._to_df(discrete_covariates, idx, "discrete_covariates")
         cont = self._to_df(continuous_covariates, idx, "continuous_covariates")
 
@@ -466,7 +491,7 @@ class ComBatModel:
         elif method == "chen":
             return self._transform_chen(X, batch, disc, cont)
         else:
-            raise ValueError(f"Unknown method: {method}")
+            raise ValueError(f"Unknown method: {method}.")
 
     def _transform_johnson(
         self,
@@ -485,7 +510,7 @@ class ComBatModel:
             if not idx.any():
                 continue
             if self.reference_batch is not None and lvl == self.reference_batch:
-                X_adj.loc[idx] = X.loc[idx].values  # untouched
+                X_adj.loc[idx] = X.loc[idx].values
                 continue
 
             g = self._gamma_star[i]
@@ -505,18 +530,28 @@ class ComBatModel:
         cont: Optional[pd.DataFrame],
     ) -> pd.DataFrame:
         """Fortin transform implementation."""
-        batch_dummies = pd.get_dummies(batch, drop_first=False)[self._batch_levels]
-        parts: list[pd.DataFrame] = [batch_dummies]
+        batch_dummies = pd.get_dummies(batch, drop_first=False).astype(float)[self._batch_levels]
+        if self.reference_batch is not None:
+            batch_dummies.loc[:, self.reference_batch] = 1.0
+
+        parts = [batch_dummies]
         if disc is not None:
-            parts.append(pd.get_dummies(disc.astype("category"), drop_first=True))
+            parts.append(
+                pd.get_dummies(
+                    disc.astype("category"), drop_first=True
+                ).astype(float)
+            )
         if cont is not None:
-            parts.append(cont)
+            parts.append(cont.astype(float))
 
-        design = pd.concat(parts, axis=1).astype(float).values
+        design = pd.concat(parts, axis=1).values
 
         X_np = X.values
-        stand_mean = self._grand_mean.values + design[:, self._n_batch:] @ self._beta_hat_nonbatch
-        Xs = (X_np - stand_mean) / np.sqrt(self._pooled_var.values)
+        stand_mu = (
+            self._grand_mean.values +
+            design[:, self._n_batch:] @ self._beta_hat_nonbatch
+        )
+        Xs = (X_np - stand_mu) / np.sqrt(self._pooled_var.values)
 
         for i, lvl in enumerate(self._batch_levels):
             idx = batch == lvl
@@ -533,8 +568,11 @@ class ComBatModel:
             else:
                 Xs[idx] = (Xs[idx] - g) / np.sqrt(d)
 
-        X_adj = Xs * np.sqrt(self._pooled_var.values) + stand_mean
-        return pd.DataFrame(X_adj, index=X.index, columns=X.columns)
+        X_adj = (
+            Xs * np.sqrt(self._pooled_var.values) +
+            stand_mu
+        )
+        return pd.DataFrame(X_adj, index=X.index, columns=X.columns, dtype=float)
     
     def _transform_chen(
         self,
@@ -568,7 +606,7 @@ class ComBatModel:
 
 
 class ComBat(BaseEstimator, TransformerMixin):
-    """Pipeline‑friendly wrapper around `ComBatModel`.
+    """Pipeline-friendly wrapper around `ComBatModel`.
 
     Stores batch (and optional covariates) passed at construction and
     appropriately uses them for separate `fit` and `transform`.

{combatlearn-0.1.1 → combatlearn-0.1.2/combatlearn.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: combatlearn
-Version: 0.1.1
+Version: 0.1.2
 Summary: Batch-effect harmonization for machine learning frameworks.
 Author-email: Ettore Rocchi <ettoreroc@gmail.com>
 License: MIT License
@@ -42,10 +42,10 @@ Dynamic: license-file
 
 # **combatlearn**
 
-[![Python versions](https://img.shields.io/badge/python-3.10%20%7C%203.11%20%7C%203.12-blue?logo=python)](https://img.shields.io/badge/python-3.10%20%7C%203.11%20%7C%203.12-blue?logo=python)
+[![Python versions](https://img.shields.io/badge/python-%3E%3D3.10-blue?logo=python)](https://www.python.org/)
 [![Test](https://github.com/EttoreRocchi/combatlearn/actions/workflows/test.yaml/badge.svg)](https://github.com/EttoreRocchi/combatlearn/actions/workflows/test.yaml)
 [![PyPI Downloads](https://static.pepy.tech/badge/combatlearn)](https://pepy.tech/projects/combatlearn)
-[![PyPI version](https://badge.fury.io/py/combatlearn.svg)](https://pypi.org/project/combatlearn/)
+[![PyPI Version](https://img.shields.io/pypi/v/combatlearn?cacheSeconds=300)](https://pypi.org/project/combatlearn/)
 [![License](https://img.shields.io/github/license/EttoreRocchi/combatlearn)](https://github.com/EttoreRocchi/combatlearn/blob/main/LICENSE)
 
 <div align="center">
@@ -56,7 +56,7 @@ Dynamic: license-file
 
 **Three methods**:
 - `method="johnson"` - classic ComBat (Johnson _et al._, 2007)
-- `method="fortin"` - covariate-aware ComBat (Fortin _et al._, 2018)
+- `method="fortin"` - neuroComBat (Fortin _et al._, 2018)
 - `method="chen"` - CovBat (Chen _et al._, 2022)
 
 ## Installation

combatlearn-0.1.2/combatlearn.egg-info/SOURCES.txt

@@ -0,0 +1,11 @@
+LICENSE
+README.md
+pyproject.toml
+combatlearn/__init__.py
+combatlearn/combat.py
+combatlearn.egg-info/PKG-INFO
+combatlearn.egg-info/SOURCES.txt
+combatlearn.egg-info/dependency_links.txt
+combatlearn.egg-info/requires.txt
+combatlearn.egg-info/top_level.txt
+tests/test_combat.py

{combatlearn-0.1.1 → combatlearn-0.1.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "combatlearn"
-version = "0.1.1"
+version = "0.1.2"
 description = "Batch-effect harmonization for machine learning frameworks."
 authors = [{name="Ettore Rocchi", email="ettoreroc@gmail.com"}]
 requires-python = ">=3.10"
@@ -31,5 +31,5 @@ classifiers = [
 ]
 
 [tool.setuptools.packages.find]
-where = ["src"]
+where = ["."]
 include = ["combatlearn*"]

combatlearn-0.1.1/src/combatlearn.egg-info/SOURCES.txt

@@ -1,11 +0,0 @@
-LICENSE
-README.md
-pyproject.toml
-src/combatlearn/__init__.py
-src/combatlearn/combat.py
-src/combatlearn.egg-info/PKG-INFO
-src/combatlearn.egg-info/SOURCES.txt
-src/combatlearn.egg-info/dependency_links.txt
-src/combatlearn.egg-info/requires.txt
-src/combatlearn.egg-info/top_level.txt
-tests/test_combat.py