Merge pull request #89 from spotify/Release-3.0.1

Added reference level point estimate to hover box of difference plots
spotify · Apr 21, 2023 · af091a3 · af091a3
2 parents 1b43ec4 + c821056
commit af091a3
Show file tree

Hide file tree

Showing 5 changed files with 41 additions and 7 deletions.
diff --git a/HISTORY.rst b/HISTORY.rst
@@ -2,6 +2,10 @@
 History
 =======
 
+3.0.1 (2023-04-20)
+------------------
+* Added reference level point estimate to the hover box of difference plots
+
 3.0.0 (2023-03-24)
 ------------------
 * Dropped support for python 3.6

diff --git a/README.md b/README.md
@@ -2,11 +2,11 @@ Spotify Confidence
 ========
 
 ![Status](https://img.shields.io/badge/Status-Beta-blue.svg)
-![Latest release](https://img.shields.io/badge/release-3.0.0-green.svg "Latest release: 3.0.0")
-![Python](https://img.shields.io/badge/Python-3.6-blue.svg "Python")
+![Latest release](https://img.shields.io/badge/release-3.0.1-green.svg "Latest release: 3.0.1")
 ![Python](https://img.shields.io/badge/Python-3.7-blue.svg "Python")
 ![Python](https://img.shields.io/badge/Python-3.8-blue.svg "Python")
 ![Python](https://img.shields.io/badge/Python-3.9-blue.svg "Python")
+![Python](https://img.shields.io/badge/Python-3.10-blue.svg "Python")
 
 Python library for AB test analysis.
 

diff --git a/setup.cfg b/setup.cfg
@@ -1,6 +1,6 @@
 [metadata]
 name = spotify-confidence
-version = 3.0.0
+version = 3.0.1
 author = Per Sillren
 author_email = [email protected]
 description = Package for calculating and visualising confidence intervals, e.g. for A/B test analysis.
@@ -24,7 +24,7 @@ install_requires =
     scipy>=1.6.0,<1.8.0
     pandas>=1.2.0,<2.0.0
     statsmodels>=0.13.0,<1.0.0
-    chartify>=4.0.2
+    chartify>=4.0.3
     ipywidgets>=8.0.0
 
 [options.packages.find]

diff --git a/spotify_confidence/analysis/frequentist/chartify_grapher.py b/spotify_confidence/analysis/frequentist/chartify_grapher.py
@@ -32,6 +32,7 @@
 )
 from ..constants import (
     POINT_ESTIMATE,
+    ORIGINAL_POINT_ESTIMATE,
     DIFFERENCE,
     CI_LOWER,
     CI_UPPER,
@@ -43,6 +44,7 @@
     NIM,
     NIM_TYPE,
     PREFERENCE,
+    SFX1,
 )
 from ...chartgrid import ChartGrid
 
@@ -184,7 +186,7 @@ def _ordinal_difference_plot(
 
         y_axis_label = self._get_difference_plot_label(absolute)
         ch = self._ordinal_plot(
-            "difference",
+            DIFFERENCE,
             difference_df,
             groupby=None,
             level_name="",
@@ -381,6 +383,8 @@ def _ordinal_plot(
             y_column=center_name,
             color_column=colors,
         )
+        # Also plot transparent circles, just to be able to show hover box
+        ch.figure.line(source=df, x=self._ordinal_group_column, y=center_name, name="center", line_alpha=0)
         ch.style.color_palette.reset_palette_order()
         ch.plot.area(
             data_frame=(
@@ -403,6 +407,15 @@ def _ordinal_plot(
                 line_dash="dashed",
                 line_width=1,
             )
+            # Also plot named transparent line, just to be able to show hover box
+            ch.figure.line(
+                source=df.sort_values(self._ordinal_group_column),
+                x=self._ordinal_group_column,
+                y=NULL_HYPOTHESIS,
+                line_width=3,
+                line_alpha=0,
+                name="nim",
+            )
         ch.axes.set_yaxis_label(y_axis_label)
         ch.axes.set_xaxis_label(self._ordinal_group_column)
         ch.set_source_label("")
@@ -553,9 +566,11 @@ def add_ci_to_chart_datasources(
                 data["color"] = np.array(df[group_col][index])
             if DIFFERENCE in data.keys() or NULL_HYPOTHESIS in data.keys():
                 index = data["index"]
+                data["reference_level"] = np.array(df["level_1"][index])
                 data[DIFFERENCE] = np.array(df[DIFFERENCE][index])
                 data["p_value"] = np.array(df[P_VALUE][index])
                 data["adjusted_p"] = np.array(df[ADJUSTED_P][index])
+                data["reference_level_avg"] = np.array(df[ORIGINAL_POINT_ESTIMATE + SFX1][index])
                 if NULL_HYPOTHESIS in df.columns:
                     data["null_hyp"] = np.array(df[NULL_HYPOTHESIS][index])
 
@@ -580,6 +595,13 @@ def add_tools(
             absolute=absolute,
             extra_zeros=2,
         )
+        axis_format_reference_level, _, _ = axis_format_precision(
+            numbers=concat(
+                [df[LOWER], df[center_name], df[UPPER], df[NULL_HYPOTHESIS] if NULL_HYPOTHESIS in df.columns else None]
+            ),
+            absolute=True,
+            extra_zeros=2,
+        )
         ordinal_tool_tip = [] if not ordinal else [(self._ordinal_group_column, f"@{self._ordinal_group_column}")]
         p_value_tool_tip = (
             (
@@ -590,8 +612,14 @@ def add_tools(
             else []
         )
         nim_tool_tip = [("null hypothesis", f"@null_hyp{{{axis_format}}}")] if NULL_HYPOTHESIS in df.columns else []
+        reference_level_tool_tip = (
+            [("reference level", f"@reference_level: @reference_level_avg{{{axis_format_reference_level}}}")]
+            if "level_1" in df.columns
+            else []
+        )
         tooltips = (
             [("group", "@color")]
+            + reference_level_tool_tip
             + ordinal_tool_tip
             + [(f"{center_name}", f"@{center_name}{{{axis_format}}}")]
             + [
@@ -603,7 +631,7 @@ def add_tools(
             + p_value_tool_tip
             + nim_tool_tip
         )
-        lines_with_hover = [] if ordinal else ["center", "nim"]
+        lines_with_hover = ["center", "nim"]
         renderers = [r for r in chart.figure.renderers if r.name in lines_with_hover]
         hover = tools.HoverTool(tooltips=tooltips, renderers=renderers)
 

diff --git a/spotify_confidence/analysis/frequentist/experiment.py b/spotify_confidence/analysis/frequentist/experiment.py
@@ -200,6 +200,7 @@ def difference_plot(
             groupby=groupby,
             non_inferiority_margins=non_inferiority_margins,
             final_expected_sample_size_column=final_expected_sample_size_column,
+            verbose=True,
         )
         chartgrid = self._confidence_grapher.plot_difference(
             difference_df, absolute, groupby, non_inferiority_margins, use_adjusted_intervals, split_plot_by_groups
@@ -217,7 +218,7 @@ def differences_plot(
         split_plot_by_groups: bool = False,
     ) -> ChartGrid:
         difference_df = self.differences(
-            levels, absolute, groupby, non_inferiority_margins, final_expected_sample_size_column
+            levels, absolute, groupby, non_inferiority_margins, final_expected_sample_size_column, verbose=True
         )
         chartgrid = self._confidence_grapher.plot_differences(
             difference_df, absolute, groupby, non_inferiority_margins, use_adjusted_intervals, split_plot_by_groups
@@ -242,6 +243,7 @@ def multiple_difference_plot(
             level_as_reference=level_as_reference,
             non_inferiority_margins=non_inferiority_margins,
             final_expected_sample_size_column=final_expected_sample_size_column,
+            verbose=True,
         )
         chartgrid = self._confidence_grapher.plot_multiple_difference(
             difference_df,