Merge pull request #369 from bashtage/improve-error

ENH: Improve error for unequal data size

Author: bashtage

ci/azure_template_posix.yml

@@ -42,8 +42,15 @@ jobs:
       python39_no_cython:
         python.version: '3.9'
         SETUP_OPTIONS: '--no-binary'
-      python39_latest:
+      python39_recent:
         python.version: '3.9'
+        NUMPY: 1.20.3
+        SCIPY: 1.6.3
+        PANDAS: 1.2.5
+        STATSMODELS: 0.12.2
+        XARRAY: 0.18.2
+      python310_latest:
+        python.version: '3.10'
     maxParallel: 10
 
   steps:

doc/source/changes.rst

@@ -1,7 +1,14 @@
 Change Log
 ==========
+
 Version 4.25
 ------------
+* Clarified the null in the F-statistic
+* Improved the error message when dependent and exog have different numbers
+  of observations
+* Added formulaic as the preferred formula parser
+* Fixed a bug in :class:`~linearmodels.panel.covariance.ACCovariance` estimator where the number of observations
+  was incorrectly overwritten
 * Fixed a bug in :func:`linearmodels.panel.results.PanelResults.corr_squared_between`,
   :func:`linearmodels.panel.results.PanelResults.corr_squared_overall`, and
   :func:`linearmodels.panel.results.PanelResults.corr_squared_within` where the correlation

linearmodels/panel/model.py

@@ -420,7 +420,10 @@ def _validate_data(self) -> None:
         w = self._w = cast(Float64Array, self.weights.values2d)
         if y.shape[0] != x.shape[0]:
             raise ValueError(
-                "dependent and exog must have the same number of " "observations."
+                "dependent and exog must have the same number of "
+                "observations. The number of observations in dependent "
+                f"is {y.shape[0]}, and the number of observations in exog "
+                f"is {x.shape[0]}."
             )
         if y.shape[0] != w.shape[0]:
             raise ValueError(

linearmodels/tests/panel/_utility.py

@@ -1,12 +1,15 @@
+from __future__ import annotations
+
 import numpy as np
 from numpy.linalg import lstsq
-from numpy.random import standard_normal
+from numpy.random import RandomState, standard_normal
 from numpy.testing import assert_allclose
 from pandas import Categorical, DataFrame, date_range, get_dummies
 from pandas.testing import assert_frame_equal, assert_series_equal
 
 from linearmodels.panel.data import PanelData
 from linearmodels.shared.utility import AttrDict, panel_to_frame
+from linearmodels.typing import Literal
 
 try:
     import xarray  # noqa: F401
@@ -64,13 +67,13 @@ def lsdv(
 
 
 def generate_data(
-    missing,
-    datatype,
-    const=False,
-    ntk=(971, 7, 5),
-    other_effects=0,
-    rng=None,
-    num_cats=4,
+    missing: bool,
+    datatype: Literal["pandas", "xarray", "numpy"],
+    const: bool = False,
+    ntk: tuple[int, int, int] = (971, 7, 5),
+    other_effects: int = 0,
+    rng: RandomState | None = None,
+    num_cats: int = 4,
 ):
     if rng is None:
         np.random.seed(12345)

linearmodels/tests/panel/test_pooled_ols.py

@@ -250,3 +250,9 @@ def test_alt_cov(data, cov_type):
     else:
         cov_name = "Driscoll-Kraay"
     assert cov_name in str(res.summary)
+
+
+def test_uneuqal_samples():
+    data = generate_data(False, "pandas")
+    with pytest.raises(ValueError, match="dependent and exog must have"):
+        PooledOLS(data.y.iloc[::2], data.x)

pyproject.toml

@@ -6,11 +6,12 @@ requires = [
   "numpy==1.15.4; python_version=='3.7'",
   "numpy==1.17.5; python_version=='3.8'",
   "numpy==1.19.5; python_version=='3.9'",
-  "numpy; python_version>'3.9'",
-  "Cython>=0.29.22"]
+  "numpy==1.21.4; python_version=='3.10'",
+  "numpy; python_version>'3.10'",
+  "Cython>=0.29.24"]
 
 [tool.black]
-target-version = ['py37', 'py38']
+target-version = ['py37', 'py38', 'py39']
 exclude = '''
 (
     \.egg
@@ -22,4 +23,4 @@ exclude = '''
   | build
   | dist
 )
-'''
+'''