minor fixes to outlier detection tso unit test (#8623)

jwst/outlier_detection/tests/test_outlier_detection.py

@@ -346,18 +346,10 @@ def test_outlier_step_weak_cr_tso(exptype, tsovisit):
     im = we_many_sci(
         numsci=numsci, background=bkg, sigma=sig, signal=signal, exptype=exptype, tsovisit=tsovisit
     )
-    cube_data = np.array([i.data for i in im])
-    cube_err = np.array([i.err for i in im])
-    cube_dq = np.array([i.dq for i in im])
-    cube_var_noise = np.array([i.var_rnoise for i in im])
-    cube = datamodels.CubeModel(data=cube_data, err=cube_err, dq=cube_dq, var_noise=cube_var_noise)
-
-    # update metadata of cube to match the first image
-    cube.meta = im[0].meta
 
     # Drop a weak CR on the science array
     cr_timestep = 5
-    cube.data[cr_timestep, 12, 12] = bkg + sig * 10
+    im[cr_timestep].data[12, 12] = bkg + sig * 10
 
     # make time variability that has larger total amplitude than
     # the CR signal but deviations frame-by-frame are smaller
@@ -366,10 +358,20 @@ def test_outlier_step_weak_cr_tso(exptype, tsovisit):
         model.data[7, 7] += real_time_variability[i]
         model.err[7, 7] = np.sqrt(sig ** 2 + model.data[7, 7])
 
+    cube_data = np.array([i.data for i in im])
+    cube_err = np.array([i.err for i in im])
+    cube_dq = np.array([i.dq for i in im])
+    cube_var_noise = np.array([i.var_rnoise for i in im])
+    cube = datamodels.CubeModel(data=cube_data, err=cube_err, dq=cube_dq, var_noise=cube_var_noise)
+
+    # update metadata of cube to match the first image
+    cube.meta = im[0].meta
+
     result = OutlierDetectionStep.call(cube, rolling_window_width=rolling_window_width)
 
     # Make sure nothing changed in SCI array
-    np.testing.assert_allclose(cube.data, result.data)
+    for i, model in enumerate(im):
+        np.testing.assert_allclose(model.data, result.data[i])
 
     # Verify source is not flagged
     assert np.all(result.dq[:, 7, 7] == datamodels.dqflags.pixel["GOOD"])