diff --git a/ioos_qc/qartod.py b/ioos_qc/qartod.py
index 7142ad8..a9b4feb 100644
--- a/ioos_qc/qartod.py
+++ b/ioos_qc/qartod.py
@@ -491,7 +491,7 @@ def spike_test(inp: Sequence[N],
     # Apply different method
     if method == 'average':
         # Calculate the average of n-2 and n
-        ref = np.zeros(inp.size, dtype=np.float64)
+        ref = np.ma.zeros(inp.size, dtype=np.float64)
         ref[1:-1] = (inp[0:-2] + inp[2:]) / 2
         ref = np.ma.masked_invalid(ref)
 
diff --git a/tests/test_qartod.py b/tests/test_qartod.py
index 1d4fb2e..51c2bcc 100644
--- a/tests/test_qartod.py
+++ b/tests/test_qartod.py
@@ -943,7 +943,7 @@ def test_spike_masked(self):
 
         # First and last elements should always be good data, unless someone
         # has set a threshold to zero.
-        expected = [2, 4, 4, 4, 1, 3, 1, 1, 9, 9, 4, 4, 4, 9]
+        expected = [2, 4, 4, 4, 1, 3, 1, 9, 9, 9, 4, 4, 9, 9]
 
         inputs = [
             arr,
@@ -995,7 +995,7 @@ def test_spike_methods(self):
         inp = [3, 4.99, 5, 6, 8, 6, 6, 6.75, 6, 6, 5.3, 6, 6, 9, 5, None, 4, 4]
         suspect_threshold = .5
         fail_threshold = 1
-        average_method_expected = [2, 3, 1, 1, 4, 3, 1, 3, 1, 1, 3, 1, 4, 4, 4, 9, 9, 2]
+        average_method_expected = [2, 3, 1, 1, 4, 3, 1, 3, 1, 1, 3, 1, 4, 4, 9, 9, 9, 2]
         diff_method_expected = [2, 1, 1, 1, 4, 1, 1, 3, 1, 1, 3, 1, 1, 4, 9, 9, 9, 2]
 
         # Test average method
@@ -1052,8 +1052,8 @@ def test_spike_test_bad_method(self):
 
     def test_spike_test_inputs(self):
         inp = [3, 4.99, 5, 6, 8, 6, 6, 6.75, 6, 6, 5.3, 6, 6, 9, 5, None, 4, 4]
-        expected_suspect_only = [2, 3, 1, 1, 3, 3, 1, 3, 1, 1, 3, 1, 3, 3, 3, 9, 9, 2]
-        expected_fail_only = [2, 1, 1, 1, 4, 1, 1, 1, 1, 1, 1, 1, 4, 4, 4, 9, 9, 2]
+        expected_suspect_only = [2, 3, 1, 1, 3, 3, 1, 3, 1, 1, 3, 1, 3, 3, 9, 9, 9, 2]
+        expected_fail_only = [2, 1, 1, 1, 4, 1, 1, 1, 1, 1, 1, 1, 4, 4, 9, 9, 9, 2]
         suspect_threshold = .5
         fail_threshold = 1