process-improve 1.75.2__tar.gz → 1.77.0__tar.gz

{process_improve-1.75.2 → process_improve-1.77.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: process-improve
-Version: 1.75.2
+Version: 1.77.0
 Summary: Designed Experiments; Latent Variables (PCA, PLS, multivariate methods with missing data); Process Monitoring; Batch data analysis.
 Keywords: Designed Experiments,Latent Variables,PCA,PLS,Multivariate Data Analysis,Batch data analysis
 Author: Kevin Dunn

{process_improve-1.75.2 → process_improve-1.77.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "process-improve"
-version = "1.75.2"
+version = "1.77.0"
 description = "Designed Experiments; Latent Variables (PCA, PLS, multivariate methods with missing data); Process Monitoring; Batch data analysis."
 readme = "README.md"
 license = "MIT"

{process_improve-1.75.2 → process_improve-1.77.0}/pyproject.toml.orig

@@ -1,6 +1,6 @@
 [project]
 name = "process-improve"
-version = "1.75.2"
+version = "1.77.0"
 description = 'Designed Experiments; Latent Variables (PCA, PLS, multivariate methods with missing data); Process Monitoring; Batch data analysis.'
 readme = "README.md"
 license = "MIT"

{process_improve-1.75.2 → process_improve-1.77.0}/src/process_improve/_linalg.py

@@ -70,7 +70,14 @@ def safe_inverse(
     if is_singular(a, cond_limit=cond_limit):
         raise np.linalg.LinAlgError(
             f"{what} is singular or ill-conditioned (condition number exceeds "
-            f"{cond_limit:.2e}); cannot invert reliably. Reduce the number of "
-            "components/terms or remove collinear columns."
+            f"{cond_limit:.2e}); cannot invert reliably. Likely causes, in the "
+            "order worth checking: exactly or near-exactly collinear columns "
+            "(a duplicated variable, a column that is a sum of others, a "
+            "one-hot set that includes every level); a constant or all-missing "
+            "column; a component count that has exhausted the variance left in "
+            "the deflated block; or values spanning so many orders of magnitude "
+            "that the cross-product loses precision (centre and scale first). "
+            "Reduce the number of components/terms or remove the offending "
+            "columns."
         )
     return np.linalg.inv(a)

process_improve-1.77.0/src/process_improve/chemistry/__init__.py

@@ -0,0 +1,57 @@
+"""(c) Kevin Dunn, 2010-2026. MIT License.
+
+Preprocessing a product-by-compound block for a sensory-to-chemistry analysis.
+
+One row per product, one column per compound, values being concentrations or
+integrated peak areas. The pipeline is four steps in a fixed order, trim then
+transform then centre then scale, and each fitting step has an
+``apply_fitted_*`` partner so held-out rows can be preprocessed with training
+constants alone:
+
+.. code-block:: python
+
+    from process_improve.chemistry import (
+        apply_transform,
+        center_and_scale,
+        classify_zero_states,
+        normalisation_check,
+        trim_by_prevalence,
+    )
+
+    states = classify_zero_states(chem, lod={"linalool": 0.02})
+    totals, outside = normalisation_check(chem)
+    kept, dropped, presence = trim_by_prevalence(chem, min_nonzero=3)
+    transformed, applied = apply_transform(kept, lod={"linalool": 0.02})
+    scaled, constants = center_and_scale(transformed, presence[kept.columns])
+
+See :mod:`process_improve.chemistry.preprocessing` for the reasoning behind
+each step.
+"""
+
+from process_improve.chemistry.preprocessing import (
+    SCALING_METHODS,
+    TRANSFORM_RULES,
+    ZERO_STATES,
+    apply_fitted_center_scale,
+    apply_fitted_transform,
+    apply_transform,
+    center_and_scale,
+    choose_transform,
+    classify_zero_states,
+    normalisation_check,
+    trim_by_prevalence,
+)
+
+__all__ = [
+    "SCALING_METHODS",
+    "TRANSFORM_RULES",
+    "ZERO_STATES",
+    "apply_fitted_center_scale",
+    "apply_fitted_transform",
+    "apply_transform",
+    "center_and_scale",
+    "choose_transform",
+    "classify_zero_states",
+    "normalisation_check",
+    "trim_by_prevalence",
+]