Merge pull request #92 from DHI/misc_fix

Misc fix

docs/simple_compare.rst

@@ -1,66 +1,80 @@
-.. _simple_compare:
-
-Simple time series comparison
-#############################
-
-If all you need to do is to compare two point time series, the workflow is 
-very simple and described below. The general many-to-many comparison is decribed 
-in the `getting started guide <getting_started.html>`_.
-
-
-Workflow
-********
-
-The simplified fmskill workflow consists of these four steps:
-
-#. Specify **model result**
-#. Specify **observation**
-#. **compare()**
-#. Analysis and plotting
-
-
-1. Specify model result
-=======================
-
-The model result can be either a dfs0 or a DataFrame. It needs to have a single item only.
-
-.. code-block:: python
-
-    from mikeio import Dfs0
-    fn_mod = '../tests/testdata/SW/ts_storm_4.dfs0'
-    df_mod = Dfs0(fn_mod).read(items=0).to_dataframe()
-
-
-2. Specify Observation
-======================
-The observation can be either a dfs0, a DataFrame or a PointObservation object. 
-It needs to have a single item only.
-
-.. code-block:: python
-
-    fn_obs = '../tests/testdata/SW/eur_Hm0.dfs0'
-
-
-3. compare()
-============
-The `compare() <api.html#fmskill.connection.compare>`_ method will interpolate the modelresult to the time of the observation
-and return an object that can be used for analysis and plotting
-
-.. code-block:: python
-
-    import fmskill
-    c = fmskill.compare(fn_obs, df_mod)
-
-
-4. Analysis and plotting
-========================
-
-The returned `PointComparer <api.html#fmskill.comparison.PointComparer>`_ can make
-scatter plots, skill assessment, time series plots etc.
-
-
-.. code-block:: python
-
-    c.plot_timeseries()
-    c.skill()
-    c.scatter()
+.. _simple_compare:
+
+Simple time series comparison
+#############################
+
+If all you need to do is to compare two point time series, the workflow is 
+very simple and described below. The general many-to-many comparison is decribed 
+in the `getting started guide <getting_started.html>`_.
+
+
+Workflow
+********
+
+The simplified fmskill workflow consists of these four steps:
+
+#. Specify **model result**
+#. Specify **observation**
+#. **compare()**
+#. Analysis and plotting
+
+
+1. Specify model result
+=======================
+
+The model result can be either a dfs0 or a DataFrame.
+
+.. code-block:: python
+
+    from mikeio import Dfs0
+    fn_mod = '../tests/testdata/SW/ts_storm_4.dfs0'
+
+
+2. Specify Observation
+======================
+The observation can be either a dfs0, a DataFrame or a PointObservation object. 
+
+.. code-block:: python
+
+    fn_obs = '../tests/testdata/SW/eur_Hm0.dfs0'
+
+
+3. compare()
+============
+The `compare() <api.html#fmskill.connection.compare>`_ method will interpolate the modelresult to the time of the observation
+and return an object that can be used for analysis and plotting
+
+.. code-block:: python
+
+    import fmskill
+    c = fmskill.compare(fn_obs, fn_mod, mod_item=0)
+
+
+4. Analysis and plotting
+========================
+
+The returned `PointComparer <api.html#fmskill.comparison.PointComparer>`_ can make
+scatter plots, skill assessment, time series plots etc.
+
+
+.. code-block:: python
+
+    >>> c.plot_timeseries()
+
+.. image:: images/ts_plot.png
+
+
+.. code-block:: python
+
+    >>> c.scatter()
+
+.. image:: images/scatter_plot.png
+
+.. code-block:: python
+
+    >>> c.skill()
+                n     bias      rmse     urmse       mae        cc        si        r2
+    observation
+    eur_Hm0      66  0.05321  0.229957  0.223717  0.177321  0.967972  0.081732  0.929005
+
+

fmskill/comparison.py

@@ -1445,6 +1445,27 @@ def __init__(self, observation, modeldata):
     def plot_timeseries(
         self, title=None, ylim=None, figsize=None, backend="matplotlib", **kwargs
     ):
+        """Timeseries plot showing compared data: observation vs modelled
+
+        Parameters
+        ----------
+        title : str, optional
+            plot title, by default None
+        ylim : tuple, optional
+            plot range for the model (ymin, ymax), by default None
+        figsize : (float, float), optional
+            figure size, by default None
+        backend : str, optional
+            use "plotly" (interactive) or "matplotlib" backend, by default "matplotlib"backend:
+
+        Examples
+        ------
+        >>> comparer.plot_timeseries()
+        >>> comparer.plot_timeseries(title="")
+        >>> comparer.plot_timeseries(ylim=[0,6])
+        >>> comparer.plot_timeseries(backend="plotly")
+        >>> comparer.plot_timeseries(backend="plotly", showlegend=False)
+        """
 
         if title is None:
             title = self.name

fmskill/connection.py

@@ -75,6 +75,8 @@ def _parse_model(mod, item=None):
 
     assert mod.shape[1] == 1  # A single item
 
+    mod.columns = ["Model"]
+
     return mod
 
 
@@ -488,9 +490,7 @@ def _get_obs_id(self, obs):
             if obs >= 0 and obs < n_con:
                 obs_id = obs
             else:
-                raise IndexError(
-                    f"connection id {obs} is out of range (0, {n_con-1})"
-                )
+                raise IndexError(f"connection id {obs} is out of range (0, {n_con-1})")
         else:
             raise KeyError("connection must be None, str or int")
         return obs_id

fmskill/metrics.py

@@ -435,8 +435,9 @@ def hit_ratio(obs: np.ndarray, model: np.ndarray, a=0.1) -> float:
     >>> hit_ratio(obs, model, a=0.15)
     1.0
     """
+    assert obs.size == model.size
 
-    return np.mean(np.abs(obs - model) < a)
+    return np.mean(np.abs(obs.ravel() - model.ravel()) < a)
 
 
 def lin_slope(obs: np.ndarray, model: np.ndarray, reg_method="ols") -> float:
@@ -449,14 +450,14 @@ def lin_slope(obs: np.ndarray, model: np.ndarray, reg_method="ols") -> float:
 
     Range: :math:`(-\\infty, \\infty )`; Best: 1
     """
-    return _linear_regression(obs, model, reg_method)[0]
+    assert obs.size == model.size
+    return _linear_regression(obs.ravel(), model.ravel(), reg_method)[0]
 
 
 def _linear_regression(
     obs: np.ndarray, model: np.ndarray, reg_method="ols"
 ) -> Tuple[float, float]:
 
-    assert obs.size == model.size
     if len(obs) == 0:
         return np.nan
 

fmskill/skill.py

@@ -357,6 +357,39 @@ def plot_bar(self, field, level=0, **kwargs):
                 kwargs["title"] = field
         return df.plot.bar(**kwargs)
 
+    def plot_barh(self, field, level=0, **kwargs):
+        """plot statistic as horizontal bar chart using pd.DataFrame.plot.barh()
+
+        Parameters
+        ----------
+        field : str
+            field (statistic) to plot e.g. "rmse"
+        level : int or str, optional
+            level to unstack, by default 0
+        kwargs : dict, optional
+            key word arguments to be passed to pd.DataFrame.plot.bar()
+            e.g. color, title, figsize, ...
+
+        Returns
+        -------
+        AxesSubplot
+
+        Examples
+        --------
+        >>> s = comparer.skill()
+        >>> s.plot_barh("rmse")
+        >>> s.plot_barh("mae", level="observation")
+        >>> s.plot_barh(field="si", title="scatter index")
+        """
+        if isinstance(self.index, pd.MultiIndex):
+            df = self.df[field].unstack(level=level)
+        else:
+            df = self.df[field]
+        if "title" not in kwargs:
+            if isinstance(field, str):
+                kwargs["title"] = field
+        return df.plot.barh(**kwargs)
+
     def plot_grid(
         self,
         field,