filter exterme outliers before plotting the boxplots

batch_analysis_analysis.py

@@ -44,11 +44,18 @@ def main():
         for idx, (col, color) in enumerate(zip(['Mean', 'Std', 'Min', 'Max', 'Count'], colors)):
             ax = axs[idx]
 
+            # Prepare the data for the current column
+            current_plot_data = plot_data[col].dropna()
+            # Remove outliers for better visualization
+            filtered_plot_data = current_plot_data[current_plot_data.between(current_plot_data.quantile(.03), current_plot_data.quantile(0.97))]
+
+            filtered_count = current_plot_data.count() - filtered_plot_data.count()
+
             # Create boxplots
-            sns.boxplot(data=plot_data[col], ax=ax, color=color, showfliers=True, width=0.4)
+            sns.boxplot(data=filtered_plot_data, ax=ax, color=color, showfliers=False, width=0.4) # type: ignore
 
             # Add individual data points
-            sns.swarmplot(data=plot_data[col], ax=ax, color='black', size=3, alpha=0.6)
+            sns.swarmplot(data=filtered_plot_data, ax=ax, color='black', size=3, alpha=0.6) # type: ignore
 
             # Set labels and title
             ax.set_title(f'{col} Distribution', fontsize=14, fontweight='bold')
@@ -56,13 +63,22 @@ def main():
             ax.set_xlabel('') # No x-label needed
             ax.set_ylabel('Latency (ms)' if col != 'Count' else 'Count', fontsize=12)
 
-            # Calculate statistics of the statistics
+            # Calculate statistics of the statistics - here based on the original data with outliers!
             data_values = plot_data[col]
+            first_line_length = len(f"Mean: {data_values.mean():.2f}")
+            second_line_length = len(f"Std: {data_values.std():.2f}")
+            third_line_length = len(f"Min: {data_values.min():.2f}")
+            fourth_line_length = len(f"Max: {data_values.max():.2f}")
+            fivth_line_length = len(f"Filtered: {filtered_count}")
+            max_length = max(first_line_length, second_line_length, third_line_length, fourth_line_length, fivth_line_length) + 1
+            # Prepare the text for the legend
+
             stats_text = (
-                f"Mean: {data_values.mean():.2f}\n"
+                f"Mean:{' ' * (max_length - first_line_length)}{data_values.mean():.2f}\n"
-                f"Std: {data_values.std():.2f}\n"
+                f"Std:{' ' * (max_length - second_line_length)}{data_values.std():.2f}\n"
-                f"Min: {data_values.min():.2f}\n"
+                f"Min:{' ' * (max_length - third_line_length)}{data_values.min():.2f}\n"
-                f"Max: {data_values.max():.2f}"
+                f"Max:{' ' * (max_length - fourth_line_length)}{data_values.max():.2f}\n"
+                f"Filtered:{' ' * (max_length - fivth_line_length)}{filtered_count}"
             )
 
             # --- Place legend in the top right using axes fraction coordinates ---
@@ -73,6 +89,7 @@ def main():
                 verticalalignment='top',
                 horizontalalignment='right',
                 fontsize=10,
+                fontfamily='monospace',
                 bbox=dict(facecolor='white', alpha=0.9, boxstyle='round,pad=0.3', edgecolor='gray')
             )
 
@@ -86,7 +103,6 @@ def main():
         )
 
         # Save the figure with a filename that includes the chain name
-        plt.tight_layout()
         output_file = args.input.replace('.csv', f'_chain_{chain_name_fs}_analysis.png')
         plt.savefig(output_file, dpi=300)
         plt.close()