Added proper formatting to tables

mobgap/utils/df_operations.py

@@ -387,6 +387,7 @@ def apply_transformations(  # noqa: C901, PLR0912
             "This is likely due to an unexpected return type of a custom function."
             "Please ensure that the return type is a pandas Series for all custom functions."
         ) from e
+    column_names = [col_name if isinstance(col_name, tuple) else (col_name,) for col_name in column_names]
     try:
         transformation_results.columns = pd.MultiIndex.from_tuples(column_names)
     except ValueError as e:

mobgap/utils/tables.py

@@ -0,0 +1,40 @@
+"""Tools to format pandas tables for documentation and publication."""
+
+import pandas as pd
+
+
+def value_with_range(df: pd.DataFrame, value_col: str, range_col: str, precision: int = 2) -> pd.Series:
+    """Combine a value column (float) and a range column tuple(float, float) into one string.
+
+    Parameters
+    ----------
+    df
+        The DataFrame containing the columns.
+    value_col
+        The name of the column containing the value.
+    range_col
+        The name of the column containing the range.
+    precision
+        The precision to use for the value and range.
+
+    """
+    return df.apply(
+        lambda x: f"{x[value_col]:.{precision}f} [{x[range_col][0]:.{precision}f}, {x[range_col][1]:.{precision}f}]",
+        axis=1,
+    )
+
+
+class FormatTransformer:
+    """Formatting functions that can be applied to a DataFrame using :func:`~mobgap.utils.df_operations.apply_transformations`.
+
+    These functions can be applied to individual columns or to multiple columns combined
+    by using :class:`~mobgap.utils.df_operations.CustomOperation`.
+    If you want to perform styling per value, usually the built-in pandas styling functions are more appropriate.
+
+    Attributes
+    ----------
+    value_with_range
+        Combine a value column (float) and a range column tuple(float, float) into one string.
+    """  # noqa: E501
+
+    value_with_range = value_with_range

revalidation/gait_sequences/_01_analysis.py

@@ -126,7 +126,12 @@
 from functools import partial
 
 from mobgap.pipeline.evaluation import CustomErrorAggregations as A
-from mobgap.utils.df_operations import CustomOperation, apply_aggregations
+from mobgap.utils.df_operations import (
+    CustomOperation,
+    apply_aggregations,
+    apply_transformations,
+)
+from mobgap.utils.tables import FormatTransformer as F
 
 custom_aggs = [
     CustomOperation(
@@ -155,10 +160,52 @@
     ),
 ]
 
+format_transforms = [
+    *(
+        CustomOperation(
+            identifier=None,
+            function=partial(
+                F.value_with_range,
+                value_col=("mean", c),
+                range_col=("conf_intervals", c),
+            ),
+            column_name=c,
+        )
+        for c in [
+            "recall",
+            "precision",
+            "f1_score",
+            "accuracy",
+            "specificity",
+            "reference_gs_duration_s",
+            "detected_gs_duration_s",
+            "gs_absolute_duration_error_s",
+        ]
+    ),
+    CustomOperation(
+        identifier=None,
+        function=partial(
+            F.value_with_range,
+            value_col=("mean", "gs_duration_error_s"),
+            range_col=("loa", "gs_duration_error_s"),
+        ),
+        column_name="gs_duration_error_s",
+    ),
+    CustomOperation(
+        identifier=None,
+        function=partial(
+            F.value_with_range,
+            value_col=("icc", "gs_duration_s"),
+            range_col=("icc_ci", "gs_duration_s"),
+        ),
+        column_name="icc",
+    ),
+]
+
 perf_metrics_all = results.groupby(["algo", "version"]).apply(
     apply_aggregations, custom_aggs
 )
-perf_metrics_all.T
+perf_metrics_all.pipe(apply_transformations, format_transforms)
 
 # %%
 # Per Cohort
@@ -172,10 +219,10 @@
 perf_metrics_per_cohort = (
     results.groupby(["cohort", "algo", "version"])
     .apply(apply_aggregations, custom_aggs)
-    .swaplevel(axis=1)
+    .pipe(apply_transformations, format_transforms)
     .loc[cohort_order]
 )
-perf_metrics_per_cohort.T
+perf_metrics_per_cohort
 
 # %%
 # Per relevant cohort
@@ -214,7 +261,7 @@
 # %%
 perf_metrics_per_cohort.loc[
     pd.IndexSlice[low_impairment_cohorts, low_impairment_algo], :
-].reset_index("algo", drop=True).T
+].reset_index("algo", drop=True)
 
 # %%
 high_impairment_algo = "GsdIonescu"
@@ -246,4 +293,4 @@
 # %%
 perf_metrics_per_cohort.loc[
     pd.IndexSlice[high_impairment_cohorts, high_impairment_algo], :
-].reset_index("algo", drop=True).T
+].reset_index("algo", drop=True)