Merge pull request #25 from arjunsavel/test_utils_more

test PCA

src/scope/tests/test_utils.py

@@ -59,3 +59,61 @@ def test_cube_normalized_max_still_max(self):
 )
 def test_calc_limb_darkening(values, output):
     assert calc_limb_darkening(*values) == output
+
+
+class TestPca(unittest.TestCase):
+    """
+    These don't really test the fancy statistics of PCA. just assert that it's a
+    variance-reducing black box.
+    """
+
+    n_exp = 10
+    n_pix = 10
+    cube = np.random.random((n_exp, n_pix))
+
+    def test_perform_pca_shapes(self):
+        """
+        just make sure the output shapes are what we expect.
+        """
+
+        n_princ_comp = 5
+        scaled_cube, pca = perform_pca(self.cube, n_princ_comp, return_noplanet=True)
+        assert scaled_cube.shape == self.cube.shape
+        assert pca.shape == self.cube.shape
+
+    def test_perform_pca_variance_removed(self):
+        """
+        just make sure the output shapes are what we expect.
+        """
+
+        n_princ_comp = 5
+        scaled_cube, pca = perform_pca(self.cube, n_princ_comp, return_noplanet=True)
+        assert np.std(scaled_cube) < np.std(
+            pca
+        )  # the variance should be contained in the main components
+
+    def test_perform_pca_variance_removed_compared_input(self):
+        """
+        just make sure the output shapes are what we expect.
+        """
+
+        n_princ_comp = 5
+        scaled_cube, pca = perform_pca(self.cube, n_princ_comp, return_noplanet=True)
+        assert np.std(scaled_cube) < np.std(
+            self.cube
+        )  # the variance should be contained in the main components
+
+    def test_perform_pca_variance_removed_compared_input_diffcompoents(self):
+        """
+        just make sure the output shapes are what we expect.
+        """
+
+        n_princ_comp = 5
+        scaled_cube5, pca5 = perform_pca(self.cube, n_princ_comp, return_noplanet=True)
+        n_princ_comp = 10
+        scaled_cube10, pca10 = perform_pca(
+            self.cube, n_princ_comp, return_noplanet=True
+        )
+        assert np.std(scaled_cube10) < np.std(
+            scaled_cube5
+        )  # the variance should be contained in the main components

src/scope/utils.py

@@ -162,7 +162,7 @@ def perform_pca(input_matrix, n_princ_comp, return_noplanet=False):
     Inputs
     ------
         :input_matrix:
-        :n_princ_comp: number of principle components to keep
+        :n_princ_comp: number of principal components to keep
     """
     u, singular_values, vh = np.linalg.svd(
         input_matrix, full_matrices=False
@@ -218,7 +218,6 @@ def calc_exposure_time(
     plot=False,
 ):
     """
-    Todo: use the newer version of exoplanet, where the orbits are installed elsewhere.
 
     todo: refactor this into separate functions, maybe?