make CLR robust against observations with O11=O21=0

fau-klue · Aug 30, 2022 · 2e3ce92 · 2e3ce92
1 parent ffdd443
commit 2e3ce92
Show file tree

Hide file tree

Showing 4 changed files with 2,153 additions and 0 deletions.
diff --git a/association_measures/measures.py b/association_measures/measures.py
@@ -343,6 +343,9 @@ def get_poisson_ci_boundary(alpha, O11, R1, O21, R2):
     :param int R2:
     """
 
+    if O11 == O21 == 0:
+        return 0
+
     if (O11 / R1) >= (O21 / R2):
         lower = beta.ppf(alpha, O11, O21 + 1)
         boundary = max(np.log2((R2 / R1) * lower / (1 - lower)), 0)

diff --git a/tests/conftest.py b/tests/conftest.py
@@ -51,6 +51,16 @@ def zero_dataframe():
     return df
 
 
+@pytest.fixture(scope='function')
+def zero_dataframe_sig():
+    """ Sample DataFrame with lots of zeros in freq. signature notation """
+
+    df = pd.read_csv("tests/data/df-zeros-sig.tsv", index_col=0,
+                     sep="\t", quoting=3, keep_default_na=False)
+
+    return df
+
+
 @pytest.fixture(scope='function')
 def ucs_dataframe():
     """ Sample DataFrame with real data and calculations from UCS.