DOC: Fix many aliases

doc/source/conf.py

@@ -23,6 +23,15 @@
 
 # More warnings
 nitpicky = True
+nitpick_ignore = []
+
+with open("nitpick-exceptions") as nitpick_ex:
+    for line in nitpick_ex:
+        if line.strip() == "" or line.startswith("#"):
+            continue
+        dtype, target = line.split(None, 1)
+        target = target.strip()
+        nitpick_ignore.append((dtype, target))
 
 # The short X.Y version
 full_version = parse(linearmodels.__version__)
@@ -412,3 +421,9 @@
 
 autosummary_generate = True
 autoclass_content = "class"
+
+autodoc_type_aliases = {
+    "ArrayLike": "linearmodels.typing.data.ArrayLike",
+    "IntArray": "linearmodels.typing.data.IntArray",
+    "Float64Array": "linearmodels.typing.data.Float64Array",
+}

doc/source/index.rst

@@ -67,6 +67,7 @@ linearmodels
    system/index
    utility
    compatibility
+   types
    plan
    contributing
    changes

doc/source/nitpick-exceptions

@@ -0,0 +1,6 @@
+# py:class ArrayLike
+py:class IVDataLike
+# py:class IntArray
+# py:class Float64Array
+
+

doc/source/types.rst

@@ -0,0 +1,27 @@
+Type Aliases
+============
+
+.. module:: linearmodels.typing.data
+
+.. automodule:: linearmodels.typing.data
+
+.. autoclass::
+
+   ArrayLike
+   Float64Array
+   Int64Array
+   Int32Array
+   IntArray
+   BoolArray
+   AnyArray
+   Uint32Array = np.ndarray[Any, np.dtype[np.uint32]]  # pragma: no cover
+
+
+.. module:: linearmodels.typing
+
+.. automodule:: linearmodels.typing
+
+.. autoclass::
+
+   Numeric
+   OptionalNumeric

examples/panel_data-formats.ipynb

@@ -49,7 +49,7 @@
    "source": [
     "orig_mi_data = data.set_index([\"fcode\", \"year\"])\n",
     "# Subset to the relevant columns and drop missing to avoid warnings\n",
-    "mi_data = orig_mi_data[[\"lscrap\",\"hrsemp\"]]\n",
+    "mi_data = orig_mi_data[[\"lscrap\", \"hrsemp\"]]\n",
     "mi_data = mi_data.dropna(axis=0, how=\"any\")\n",
     "\n",
     "print(mi_data.head())"

linearmodels/asset_pricing/covariance.py

@@ -4,22 +4,23 @@
 
 from __future__ import annotations
 
-from numpy import empty, ndarray
+import numpy
+from numpy import empty
 from numpy.linalg import inv
 
 from linearmodels.iv.covariance import (
     KERNEL_LOOKUP,
     cov_kernel,
     kernel_optimal_bandwidth,
 )
-from linearmodels.typing import Float64Array
+import linearmodels.typing.data
 
 
 class _HACMixin:
     def __init__(self, kernel: str, bandwidth: float | None) -> None:
         self._kernel: str | None = None
         self._bandwidth: float | None = None  # pragma: no cover
-        self._moments: ndarray = empty((0,))  # pragma: no cover
+        self._moments: numpy.ndarray = empty((0,))  # pragma: no cover
         self._check_kernel(kernel)
         self._check_bandwidth(bandwidth)
 
@@ -60,7 +61,9 @@ def _check_bandwidth(self, bandwidth: float | None) -> None:
             if bandwidth < 0:
                 raise ValueError("bandwidth must be non-negative.")
 
-    def _kernel_cov(self, z: Float64Array) -> Float64Array:
+    def _kernel_cov(
+        self, z: linearmodels.typing.data.Float64Array
+    ) -> linearmodels.typing.data.Float64Array:
         nobs = z.shape[0]
         bw = self.bandwidth
         kernel = self._kernel
@@ -96,10 +99,10 @@ class HeteroskedasticCovariance:
 
     def __init__(
         self,
-        xe: Float64Array,
+        xe: linearmodels.typing.data.Float64Array,
         *,
-        jacobian: ndarray | None = None,
-        inv_jacobian: ndarray | None = None,
+        jacobian: numpy.ndarray | None = None,
+        inv_jacobian: numpy.ndarray | None = None,
         center: bool = True,
         debiased: bool = False,
         df: int = 0,
@@ -131,7 +134,7 @@ def config(self) -> dict[str, str | float]:
         return {"type": self.__class__.__name__}
 
     @property
-    def s(self) -> Float64Array:
+    def s(self) -> linearmodels.typing.data.Float64Array:
         """
         Score/moment condition covariance
 
@@ -149,7 +152,7 @@ def s(self) -> Float64Array:
         return (out + out.T) / 2
 
     @property
-    def jacobian(self) -> Float64Array:
+    def jacobian(self) -> linearmodels.typing.data.Float64Array:
         """The Jacobian"""
         if self._jac is None:
             assert self._inv_jac is not None
@@ -158,7 +161,7 @@ def jacobian(self) -> Float64Array:
         return self._jac
 
     @property
-    def inv_jacobian(self) -> Float64Array:
+    def inv_jacobian(self) -> linearmodels.typing.data.Float64Array:
         """Inverse Jacobian"""
         if self._inv_jac is None:
             assert self._jac is not None
@@ -172,7 +175,7 @@ def square(self) -> bool:
         return self._square
 
     @property
-    def cov(self) -> Float64Array:
+    def cov(self) -> linearmodels.typing.data.Float64Array:
         """
         Compute parameter covariance
 
@@ -229,10 +232,10 @@ class KernelCovariance(HeteroskedasticCovariance, _HACMixin):
 
     def __init__(
         self,
-        xe: Float64Array,
+        xe: linearmodels.typing.data.Float64Array,
         *,
-        jacobian: ndarray | None = None,
-        inv_jacobian: ndarray | None = None,
+        jacobian: numpy.ndarray | None = None,
+        inv_jacobian: numpy.ndarray | None = None,
         kernel: str | None = None,
         bandwidth: float | None = None,
         center: bool = True,
@@ -262,7 +265,7 @@ def config(self) -> dict[str, str | float]:
         return out
 
     @property
-    def s(self) -> Float64Array:
+    def s(self) -> linearmodels.typing.data.Float64Array:
         """
         Score/moment condition covariance
 
@@ -289,11 +292,15 @@ class HeteroskedasticWeight:
         Flag indicating to center the moments when computing the weights
     """
 
-    def __init__(self, moments: Float64Array, center: bool = True) -> None:
+    def __init__(
+        self, moments: linearmodels.typing.data.Float64Array, center: bool = True
+    ) -> None:
         self._moments = moments
         self._center = center
 
-    def w(self, moments: Float64Array) -> Float64Array:
+    def w(
+        self, moments: linearmodels.typing.data.Float64Array
+    ) -> linearmodels.typing.data.Float64Array:
         """
         Score/moment condition weighting matrix
 
@@ -335,7 +342,7 @@ class KernelWeight(HeteroskedasticWeight, _HACMixin):
 
     def __init__(
         self,
-        moments: Float64Array,
+        moments: linearmodels.typing.data.Float64Array,
         center: bool = True,
         kernel: str | None = None,
         bandwidth: float | None = None,
@@ -344,7 +351,9 @@ def __init__(
         _HACMixin.__init__(self, kernel, bandwidth)
         super().__init__(moments, center=center)
 
-    def w(self, moments: Float64Array) -> Float64Array:
+    def w(
+        self, moments: linearmodels.typing.data.Float64Array
+    ) -> linearmodels.typing.data.Float64Array:
         """
         Score/moment condition weighting matrix