Fix/simple statistics

darts/metrics/metrics.py

@@ -515,7 +515,9 @@ def coefficient_of_variation(
     """
 
     return (
-        100 * rmse(actual_series, pred_series, intersect) / actual_series.mean().mean()
+        100
+        * rmse(actual_series, pred_series, intersect)
+        / actual_series.pd_dataframe(copy=False).mean().mean()
     )
 
 

darts/models/forecasting/theta.py

@@ -265,7 +265,7 @@ def fit(self, series):
         self.length = len(series)
         # normalization of data
         if self.normalization:
-            self.mean = series.mean().mean()
+            self.mean = series.pd_dataframe(copy=False).mean().mean()
             raise_if_not(
                 not np.isclose(self.mean, 0),
                 "The mean value of the provided series is too close to zero to perform normalization",

darts/tests/test_timeseries.py

@@ -5,6 +5,7 @@
 import xarray as xr
 from tempfile import NamedTemporaryFile
 from unittest.mock import patch
+from scipy.stats import skew, kurtosis
 
 from darts.tests.base_test_class import DartsBaseTestClass
 from darts import TimeSeries, concatenate
@@ -1452,3 +1453,113 @@ def test_time_col_convert_garbage(self):
 
         with self.assertRaises(AttributeError):
             TimeSeries.from_dataframe(df=df, time_col="Time")
+
+
+class SimpleStatisticsTestCase(DartsBaseTestClass):
+
+    times = pd.date_range("20130101", "20130110", freq="D")
+    values = np.random.rand(10, 2, 1000)
+    ar = xr.DataArray(
+        values,
+        dims=("time", "component", "sample"),
+        coords={"time": times, "component": ["a", "b"]},
+    )
+    ts = TimeSeries(ar)
+
+    def test_mean(self):
+        for axis in range(3):
+            new_ts = self.ts.mean(axis=axis)
+            # check values
+            self.assertTrue(
+                np.isclose(
+                    new_ts._xa.values, self.values.mean(axis=axis, keepdims=True)
+                ).all()
+            )
+
+    def test_var(self):
+        for ddof in range(5):
+            new_ts = self.ts.var(ddof=ddof)
+            # check values
+            self.assertTrue(
+                np.isclose(new_ts.values(), self.values.var(ddof=ddof, axis=2)).all()
+            )
+
+    def test_std(self):
+        for ddof in range(5):
+            new_ts = self.ts.std(ddof=ddof)
+            # check values
+            self.assertTrue(
+                np.isclose(new_ts.values(), self.values.std(ddof=ddof, axis=2)).all()
+            )
+
+    def test_skew(self):
+        for bias in [True, False]:
+            new_ts = self.ts.skew(bias=bias)
+            # check values
+            self.assertTrue(
+                np.isclose(new_ts.values(), skew(self.values, axis=2, bias=bias)).all()
+            )
+
+    def test_kurtosis(self):
+        for bias in [True, False]:
+            for fisher in [True, False]:
+                new_ts = self.ts.kurtosis(bias=bias, fisher=fisher)
+                # check values
+                self.assertTrue(
+                    np.isclose(
+                        new_ts.values(),
+                        kurtosis(self.values, axis=2, bias=bias, fisher=fisher),
+                    ).all()
+                )
+
+    def test_min(self):
+        for axis in range(3):
+            new_ts = self.ts.min(axis=axis)
+            # check values
+            self.assertTrue(
+                np.isclose(
+                    new_ts._xa.values, self.values.min(axis=axis, keepdims=True)
+                ).all()
+            )
+
+    def test_max(self):
+        for axis in range(3):
+            new_ts = self.ts.max(axis=axis)
+            # check values
+            self.assertTrue(
+                np.isclose(
+                    new_ts._xa.values, self.values.max(axis=axis, keepdims=True)
+                ).all()
+            )
+
+    def test_sum(self):
+        for axis in range(3):
+            new_ts = self.ts.sum(axis=axis)
+            # check values
+            self.assertTrue(
+                np.isclose(
+                    new_ts._xa.values, self.values.sum(axis=axis, keepdims=True)
+                ).all()
+            )
+
+    def test_median(self):
+        for axis in range(3):
+            new_ts = self.ts.median(axis=axis)
+            # check values
+            self.assertTrue(
+                np.isclose(
+                    new_ts._xa.values, np.median(self.values, axis=axis, keepdims=True)
+                ).all()
+            )
+
+    def test_quantile(self):
+        for method in ["linear", "inverted_cdf"]:
+            for q in [0.01, 0.1, 0.5, 0.95]:
+                new_ts = self.ts.quantile(quantile=q, method=method)
+                # check values
+                self.assertTrue(
+                    np.isclose(
+                        new_ts.values(),
+                        np.quantile(self.values, q=q, method=method, axis=2),
+                    ).all()
+                )