update gitignore

.gitignore

@@ -6,11 +6,7 @@
 .env
 __pycache__/
 *.pyc
-
-=======
 /reference/
-*.svg
->>>>>>> Stashed changes
 # 2. Ignore virtual environments COMPLETELY
 # This must come BEFORE the unignore rule
 env*/

Data/show_plots.py

@@ -151,7 +151,7 @@ plt.figure(figsize=(10, 8))
 sns.heatmap(cm, annot=True, fmt='d', cmap='Blues',
             xticklabels=['0-1', '1-2', '2-3', '3-4', '4-5', '5-6', '6-7', '7-8', '8-9', '9-10'],
             yticklabels=['0-1', '1-2', '2-3', '3-4', '4-5', '5-6', '6-7', '7-8', '8-9', '9-10'])
-#plt.title('Confusion Matrix: Ground truth EDSS vs interferred EDSS (Categorized 0-10)')
+plt.title('Confusion Matrix: Ground truth EDSS vs interferred EDSS (Categorized 0-10)')
 plt.xlabel('LLM Generated EDSS')
 plt.ylabel('Ground Truth EDSS')
 plt.tight_layout()
@@ -168,98 +168,6 @@ print(cm)
 ##
 
 
-# %%  Confusion matrix
-import pandas as pd
-import matplotlib.pyplot as plt
-import numpy as np
-from sklearn.metrics import confusion_matrix, classification_report
-import seaborn as sns
-
-# Load your data from TSV file
-file_path = '/home/shahin/Lab/Doktorarbeit/Barcelona/Data/Join_edssandsub.tsv'
-df = pd.read_csv(file_path, sep='\t')
-
-# Replace comma with dot for numeric conversion in GT.EDSS and result.EDSS
-df['GT.EDSS'] = df['GT.EDSS'].astype(str).str.replace(',', '.')
-df['result.EDSS'] = df['result.EDSS'].astype(str).str.replace(',', '.')
-
-# Convert to float (handle invalid entries gracefully)
-df['GT.EDSS'] = pd.to_numeric(df['GT.EDSS'], errors='coerce')
-df['result.EDSS'] = pd.to_numeric(df['result.EDSS'], errors='coerce')
-
-# Drop rows where either column is NaN
-df_clean = df.dropna(subset=['GT.EDSS', 'result.EDSS'])
-
-# For confusion matrix, we need to categorize the values
-# Let's create categories up to 10 (0-1, 1-2, 2-3, ..., 9-10)
-def categorize_edss(value):
-    if pd.isna(value):
-        return np.nan
-    elif value <= 1.0:
-        return '0-1'
-    elif value <= 2.0:
-        return '1-2'
-    elif value <= 3.0:
-        return '2-3'
-    elif value <= 4.0:
-        return '3-4'
-    elif value <= 5.0:
-        return '4-5'
-    elif value <= 6.0:
-        return '5-6'
-    elif value <= 7.0:
-        return '6-7'
-    elif value <= 8.0:
-        return '7-8'
-    elif value <= 9.0:
-        return '8-9'
-    elif value <= 10.0:
-        return '9-10'
-    else:
-        return '10+'
-
-# Create categorical versions
-df_clean['GT.EDSS_cat'] = df_clean['GT.EDSS'].apply(categorize_edss)
-df_clean['result.EDSS_cat'] = df_clean['result.EDSS'].apply(categorize_edss)
-
-# Remove any NaN categories
-df_clean = df_clean.dropna(subset=['GT.EDSS_cat', 'result.EDSS_cat'])
-
-# Create confusion matrix
-cm = confusion_matrix(df_clean['GT.EDSS_cat'], df_clean['result.EDSS_cat'],
-                     labels=['0-1', '1-2', '2-3', '3-4', '4-5', '5-6', '6-7', '7-8', '8-9', '9-10'])
-
-# Plot confusion matrix
-plt.figure(figsize=(10, 8))
-ax = sns.heatmap(cm, annot=True, fmt='d', cmap='Blues',
-            xticklabels=['0-1', '1-2', '2-3', '3-4', '4-5', '5-6', '6-7', '7-8', '8-9', '9-10'],
-            yticklabels=['0-1', '1-2', '2-3', '3-4', '4-5', '5-6', '6-7', '7-8', '8-9', '9-10'])
-
-# Add legend text above the color bar
-# Get the colorbar object
-cbar = ax.collections[0].colorbar
-# Add text above the colorbar
-cbar.set_label('Number of Cases', rotation=270, labelpad=20)
-
-plt.xlabel('LLM Generated EDSS')
-plt.ylabel('Ground Truth EDSS')
-#plt.title('Confusion Matrix: Ground truth EDSS vs inferred EDSS (Categorized 0-10)')
-plt.tight_layout()
-plt.show()
-
-# Print classification report
-print("Classification Report:")
-print(classification_report(df_clean['GT.EDSS_cat'], df_clean['result.EDSS_cat']))
-
-# Print raw counts
-print("\nConfusion Matrix (Raw Counts):")
-print(cm)
-
-
-##
-
-
-
 
 # %% Classification 
 import pandas as pd
@@ -754,7 +662,7 @@ print("\nFirst few rows:")
 print(df.head())
 
 # Hardcode specific patient names
-patient_names = ['6b56865d']
+patient_names = ['113c1470']
 
 # Define the functional systems (columns to plot) - adjust based on actual column names
 functional_systems = ['EDSS', 'Visual', 'Sensory', 'Motor', 'Brainstem', 'Cerebellar', 'Autonomic', 'Bladder', 'Intellectual']
@@ -1275,6 +1183,7 @@ import matplotlib.colors as mcolors
 # --- Configuration ---
 plt.rcParams['font.family'] = 'Arial'
 data_path = '/home/shahin/Lab/Doktorarbeit/Barcelona/Data/comparison.tsv'
+color_json_path = '/home/shahin/Lab/Doktorarbeit/Barcelona/Data/functional_system_colors.json'
 figure_save_path = 'project/visuals/edss_combined_confusion_matrix_mixed.svg'
 
 # --- 1. Load the Dataset ---
@@ -1303,19 +1212,19 @@ n_categories = len(categories)
 
 def categorize_edss(value):
     if pd.isna(value): return np.nan
-    # Ensure value is float to avoid TypeError
+    idx = int(min(max(value, 0), 10) - 0.001) if value > 0 else 0
-    val = float(value)
-    idx = int(min(max(val, 0), 10) - 0.001) if val > 0 else 0
     return categories[min(idx, len(categories)-1)]
 
-# --- 4. Prepare Mixed Color Matrix with Saturation ---
+# --- 4. Prepare Mixed Color Matrix ---
 cell_system_counts = np.zeros((n_categories, n_categories, len(system_names)))
 
 for s_idx, (gt_col, res_col) in enumerate(functional_systems_to_plot):
-    # Fix: Ensure numeric conversion to avoid string comparison errors
+    # CRITICAL FIX: Convert to numeric and drop NaNs in one go
+    # 'coerce' turns non-numeric strings into NaN so they don't crash the script
     temp_df = df[[gt_col, res_col]].copy()
     temp_df[gt_col] = pd.to_numeric(temp_df[gt_col], errors='coerce')
     temp_df[res_col] = pd.to_numeric(temp_df[res_col], errors='coerce')
+
     valid_df = temp_df.dropna()
 
     for _, row in valid_df.iterrows():
@@ -1324,9 +1233,10 @@ for s_idx, (gt_col, res_col) in enumerate(functional_systems_to_plot):
         if gt_cat in category_to_index and res_cat in category_to_index:
             cell_system_counts[category_to_index[gt_cat], category_to_index[res_cat], s_idx] += 1
 
-# Create an RGBA image matrix (10x10x4)
+# Create an RGB image matrix (initially white/empty)
-rgba_matrix = np.zeros((n_categories, n_categories, 4))
+rgb_matrix = np.ones((n_categories, n_categories, 3)) 
 
+# Create an Alpha matrix for the "Total Count" intensity
 total_counts = np.sum(cell_system_counts, axis=2)
 max_count = np.max(total_counts) if np.max(total_counts) > 0 else 1
 
@@ -1334,602 +1244,53 @@ for i in range(n_categories):
     for j in range(n_categories):
         count_sum = total_counts[i, j]
         if count_sum > 0:
+            # Calculate weighted average color
             mixed_rgb = np.zeros(3)
             for s_idx, s_name in enumerate(system_names):
                 weight = cell_system_counts[i, j, s_idx] / count_sum
                 system_rgb = mcolors.to_rgb(color_map[s_name])
                 mixed_rgb += np.array(system_rgb) * weight
-
+            rgb_matrix[i, j] = mixed_rgb
-            # Set RGB channels
-            rgba_matrix[i, j, :3] = mixed_rgb
-
-            # Set Alpha channel (Saturation Effect)
-            # Using a square root scale to make lower counts more visible but still "lighter"
-            alpha = np.sqrt(count_sum / max_count)
-            # Ensure alpha is at least 0.1 so it's not invisible
-            rgba_matrix[i, j, 3] = max(0.1, alpha)
-        else:
-            # Empty cells are white
-            rgba_matrix[i, j] = [1, 1, 1, 0]
 
 # --- 5. Plotting ---
 fig, ax = plt.subplots(figsize=(12, 10))
 
-# Show the matrix
+# Display the mixed color matrix
-# Note: we use origin='lower' if you want 0-1 at the bottom,
+# We use alpha based on count to show density (optional, but recommended)
-# but confusion matrices usually have 0-1 at the top (origin='upper')
+im = ax.imshow(rgb_matrix, interpolation='nearest', origin='upper')
-im = ax.imshow(rgba_matrix, interpolation='nearest', origin='upper')
 
-# Add count labels
+# Add text labels for total counts in each cell
 for i in range(n_categories):
     for j in range(n_categories):
         if total_counts[i, j] > 0:
-            # Background brightness for text contrast
+            # Determine text color based on brightness of background
-            bg_color = rgba_matrix[i, j, :3]
+            lum = 0.2126 * rgb_matrix[i,j,0] + 0.7152 * rgb_matrix[i,j,1] + 0.0722 * rgb_matrix[i,j,2]
-            lum = 0.2126 * bg_color[0] + 0.7152 * bg_color[1] + 0.0722 * bg_color[2]
+            text_col = "white" if lum < 0.5 else "black"
-            # If alpha is low, background is effectively white, so use black text
+            ax.text(j, i, int(total_counts[i, j]), ha="center", va="center", color=text_col, fontsize=9)
-            text_col = "white" if (lum < 0.5 and rgba_matrix[i,j,3] > 0.5) else "black"
-            ax.text(j, i, int(total_counts[i, j]), ha="center", va="center",
-                    color=text_col, fontsize=10, fontweight='bold')
 
 # --- 6. Styling ---
-ax.set_xlabel('LLM Inference (EDSS Category)', fontsize=12, labelpad=10)
+ax.set_xlabel('LLM Inference (EDSS Category)', fontsize=12)
-ax.set_ylabel('Ground Truth (EDSS Category)', fontsize=12, labelpad=10)
+ax.set_ylabel('Ground Truth (EDSS Category)', fontsize=12)
-ax.set_title('Saturated Confusion Matrix\nColor = System Mixture | Opacity = Density', fontsize=14, pad=20)
+ax.set_title('Blended Confusion Matrix (Color = Weighted System Mixture)', fontsize=14, pad=20)
 
 ax.set_xticks(np.arange(n_categories))
 ax.set_xticklabels(categories)
 ax.set_yticks(np.arange(n_categories))
 ax.set_yticklabels(categories)
 
-# Remove the frame/spines for a cleaner look
-for spine in ax.spines.values():
-    spine.set_visible(False)
-
 # Custom Legend
 handles = [plt.Rectangle((0,0),1,1, color=color_map[name]) for name in system_names]
 labels = [name.replace('_', ' ').capitalize() for name in system_names]
-ax.legend(handles, labels, title='Functional Systems', loc='upper left',
+ax.legend(handles, labels, title='Functional Systems', loc='upper left', bbox_to_anchor=(1.05, 1), frameon=False)
-          bbox_to_anchor=(1.05, 1), frameon=False)
 
 plt.tight_layout()
 os.makedirs(os.path.dirname(figure_save_path), exist_ok=True)
 plt.savefig(figure_save_path, format='svg', bbox_inches='tight')
 plt.show()
 
-##
 
 
+#
 
 
 
-
-
-
-
-
-# %% Difference Plot Functional system
-
-import pandas as pd
-import matplotlib.pyplot as plt
-import json
-import os
-import numpy as np
-
-# --- Configuration ---
-# Set the font to Arial for all text in the plot, as per the guidelines
-plt.rcParams['font.family'] = 'Arial'
-
-# Define the path to your data file
-data_path = '/home/shahin/Lab/Doktorarbeit/Barcelona/Data/comparison.tsv'
-
-# Define the path to save the color mapping JSON file
-color_json_path = '/home/shahin/Lab/Doktorarbeit/Barcelona/Data/functional_system_colors.json'
-
-# Define the path to save the final figure
-figure_save_path = 'project/visuals/edss_functional_systems_comparison.svg'
-
-# --- 1. Load the Dataset ---
-try:
-    # Load the TSV file
-    df = pd.read_csv(data_path, sep='\t')
-    print(f"Successfully loaded data from {data_path}")
-    print(f"Data shape: {df.shape}")
-except FileNotFoundError:
-    print(f"Error: The file at {data_path} was not found.")
-    # Exit or handle the error appropriately
-    raise
-
-# --- 2. Define Functional Systems and Create Color Mapping ---
-# List of tuples containing (ground_truth_column, result_column)
-functional_systems_to_plot = [
-    ('GT.VISUAL_OPTIC_FUNCTIONS', 'result.VISUAL OPTIC FUNCTIONS'),
-    ('GT.CEREBELLAR_FUNCTIONS', 'result.CEREBELLAR FUNCTIONS'),
-    ('GT.BRAINSTEM_FUNCTIONS', 'result.BRAINSTEM FUNCTIONS'),
-    ('GT.SENSORY_FUNCTIONS', 'result.SENSORY FUNCTIONS'),
-    ('GT.PYRAMIDAL_FUNCTIONS', 'result.PYRAMIDAL FUNCTIONS'),
-    ('GT.AMBULATION', 'result.AMBULATION'),
-    ('GT.CEREBRAL_FUNCTIONS', 'result.CEREBRAL FUNCTIONS'),
-    ('GT.BOWEL_AND_BLADDER_FUNCTIONS', 'result.BOWEL AND BLADDER FUNCTIONS')
-]
-
-# Extract system names for color mapping and legend
-system_names = [name.split('.')[1] for name, _ in functional_systems_to_plot]
-
-# Define a professional color palette (dark blue theme)
-# This is a qualitative palette with distinct, accessible colors
-colors = [
-    '#003366', # Dark Blue
-    '#336699', # Medium Blue
-    '#6699CC', # Light Blue
-    '#99CCFF', # Very Light Blue
-    '#FF9966', # Coral
-    '#FF6666', # Light Red
-    '#CC6699', # Magenta
-    '#9966CC'  # Purple
-]
-
-# Create a dictionary mapping system names to colors
-color_map = dict(zip(system_names, colors))
-
-# Ensure the directory for the JSON file exists
-os.makedirs(os.path.dirname(color_json_path), exist_ok=True)
-
-# Save the color map to a JSON file
-with open(color_json_path, 'w') as f:
-    json.dump(color_map, f, indent=4)
-
-print(f"Color mapping saved to {color_json_path}")
-
-# --- 3. Calculate Agreement Percentages and Format Legend Labels ---
-agreement_percentages = {}
-legend_labels = {}
-
-for gt_col, res_col in functional_systems_to_plot:
-    system_name = gt_col.split('.')[1]
-
-    # Convert columns to numeric, setting errors to NaN
-    gt_numeric = pd.to_numeric(df[gt_col], errors='coerce')
-    res_numeric = pd.to_numeric(df[res_col], errors='coerce')
-
-    # Ensure we are comparing the same rows
-    common_index = gt_numeric.dropna().index.intersection(res_numeric.dropna().index)
-    gt_data = gt_numeric.loc[common_index]
-    res_data = res_numeric.loc[common_index]
-
-    # Calculate agreement percentage
-    if len(gt_data) > 0:
-        agreement = (gt_data == res_data).mean() * 100
-    else:
-        agreement = 0 # Handle case with no valid data
-
-    agreement_percentages[system_name] = agreement
-
-    # Format the system name for the legend (e.g., "VISUAL_OPTIC_FUNCTIONS" -> "Visual Optic Functions")
-    formatted_name = " ".join(word.capitalize() for word in system_name.split('_'))
-    legend_labels[system_name] = f"{formatted_name} ({agreement:.1f}%)"
-
-
- # --- 4. Robustly Prepare Error Data for Boxplot ---
-
-def safe_parse(s):
-    '''Convert to float, handling comma decimals (e.g., '3,5' → 3.5)'''
-    if pd.isna(s):
-        return np.nan
-    if isinstance(s, (int, float)):
-        return float(s)
-    # Replace comma with dot, then strip whitespace
-    s_clean = str(s).replace(',', '.').strip()
-    try:
-        return float(s_clean)
-    except ValueError:
-        return np.nan
-
-plot_data = []
-for gt_col, res_col in functional_systems_to_plot:
-    system_name = gt_col.split('.')[1]
-
-    # Parse both columns with robust comma handling
-    gt_numeric = df[gt_col].apply(safe_parse)
-    res_numeric = df[res_col].apply(safe_parse)
-
-    # Compute error (only where both are finite)
-    error = res_numeric - gt_numeric
-
-    # Create temp DataFrame
-    temp_df = pd.DataFrame({
-        'system': system_name,
-        'error': error
-    }).dropna()  # drop rows where either was unparseable
-
-    plot_data.append(temp_df)
-
-plot_df = pd.concat(plot_data, ignore_index=True)
-
-if plot_df.empty:
-    print("⚠️   Warning: No valid numeric error data to plot after robust parsing.")
-else:
-    print(f"✅ Prepared error data with {len(plot_df)} data points.")
-    # Diagnostic: show a few samples
-    print("\n📌 Sample errors by system:")
-    for sys, grp in plot_df.groupby('system'):
-        print(f"  {sys:25s}: n={len(grp)}, mean err = {grp['error'].mean():+.2f}, min = {grp['error'].min():+.2f}, max = {grp['error'].max():+.2f}")
-
-# Ensure categorical ordering
-plot_df['system'] = pd.Categorical(
-    plot_df['system'],
-    categories=[name.split('.')[1] for name, _ in functional_systems_to_plot],
-    ordered=True
-)
-
-
-# --- 5. Prepare Data for Diverging Stacked Bar Plot ---
-print("\n📊 Preparing diverging stacked bar plot data...")
-
-# Define bins for error direction
-def categorize_error(err):
-    if pd.isna(err):
-        return 'missing'
-    elif err < 0:
-        return 'underestimate'
-    elif err > 0:
-        return 'overestimate'
-    else:
-        return 'match'
-
-# Add category column (only on finite errors)
-plot_df_clean = plot_df[plot_df['error'].notna()].copy()
-plot_df_clean['category'] = plot_df_clean['error'].apply(categorize_error)
-
-# Count by system + category
-category_counts = (
-    plot_df_clean
-    .groupby(['system', 'category'])
-    .size()
-    .unstack(fill_value=0)
-    .reindex(columns=['underestimate', 'match', 'overestimate'], fill_value=0)
-)
-# Reorder systems
-category_counts = category_counts.reindex(system_names)
-
-# Prepare for diverging plot:
-# - Underestimates: plotted to the *left* (negative x)
-# - Overestimates: plotted to the *right* (positive x)
-# - Matches: centered (no width needed, or as a bar of width 0.2)
-underestimate_counts = category_counts['underestimate']
-match_counts = category_counts['match']
-overestimate_counts = category_counts['overestimate']
-
-# For diverging: left = -underestimate, right = overestimate
-left_counts = underestimate_counts
-right_counts = overestimate_counts
-
-# Compute max absolute bar height (for symmetric x-axis)
-max_bar = max(left_counts.max(), right_counts.max(), 1)
-plot_range = (-max_bar, max_bar)
-
-# X-axis positions: 0 = center, left systems to -1, -2, ..., right systems to +1, +2, ...
-n_systems = len(system_names)
-positions = np.arange(n_systems)
-left_positions = -positions - 0.5  # left-aligned underestimates
-right_positions = positions + 0.5  # right-aligned overestimates
-
-# --- 6. Create Diverging Stacked Bar Plot ---
-plt.figure(figsize=(12, 7))
-
-# Colors: diverging palette
-colors = {
-    'underestimate': '#E74C3C',  # Red (left)
-    'match': '#2ECC71',          # Green (center)
-    'overestimate': '#F39C12'    # Orange (right)
-}
-
-# Plot underestimates (left side)
-bars_left = plt.barh(
-    left_positions,
-    left_counts.values,
-    height=0.8,
-    left=0,  # starts at 0, extends left (since bars are negative width would be wrong; instead use negative values)
-    color=colors['underestimate'],
-    edgecolor='black',
-    linewidth=0.5,
-    alpha=0.9,
-    label='Underestimate'
-)
-
-# Plot overestimates (right side)
-bars_right = plt.barh(
-    right_positions,
-    right_counts.values,
-    height=0.8,
-    left=0,
-    color=colors['overestimate'],
-    edgecolor='black',
-    linewidth=0.5,
-    alpha=0.9,
-    label='Overestimate'
-)
-
-# Plot matches (center — narrow bar)
-# Use a very narrow width (0.2) centered at 0
-plt.barh(
-    positions,
-    match_counts.values,
-    height=0.2,
-    left=0,  # starts at 0, extends right
-    color=colors['match'],
-    edgecolor='black',
-    linewidth=0.5,
-    alpha=0.9,
-    label='Exact Match'
-)
-
-# ✨ Better: flip match to be centered symmetrically (left=-match/2, width=match)
-# For perfect symmetry:
-for i, count in enumerate(match_counts.values):
-    if count > 0:
-        plt.barh(
-            positions[i],
-            width=count,
-            left=-count/2,
-            height=0.25,
-            color=colors['match'],
-            edgecolor='black',
-            linewidth=0.5,
-            alpha=0.95
-        )
-
-# --- 7. Styling & Labels ---
-# Zero reference line
-plt.axvline(x=0, color='black', linestyle='-', linewidth=1.2, alpha=0.8)
-
-# X-axis: symmetric around 0
-plt.xlim(plot_range[0] - max_bar*0.1, plot_range[1] + max_bar*0.1)
-plt.xticks(rotation=0, fontsize=10)
-plt.xlabel('Count', fontsize=12)
-
-# Y-axis: system names at original positions (centered)
-plt.yticks(positions, [name.replace('_', '\n').replace('and', '&') for name in system_names], fontsize=10)
-plt.ylabel('Functional System', fontsize=12)
-
-# Title & layout
-plt.title('Diverging Error Direction by Functional System\n(Red: Underestimation | Green: Exact | Orange: Overestimation)', fontsize=13, pad=15)
-
-# Clean axes
-ax = plt.gca()
-ax.spines['top'].set_visible(False)
-ax.spines['right'].set_visible(False)
-ax.spines['left'].set_visible(False)  # We only need bottom axis
-ax.xaxis.set_ticks_position('bottom')
-ax.yaxis.set_ticks_position('none')
-
-# Grid only along x
-ax.xaxis.grid(True, linestyle=':', alpha=0.5)
-
-# Legend
-from matplotlib.patches import Patch
-legend_elements = [
-    Patch(facecolor=colors['underestimate'], edgecolor='black', label='Underestimate'),
-    Patch(facecolor=colors['match'], edgecolor='black', label='Exact Match'),
-    Patch(facecolor=colors['overestimate'], edgecolor='black', label='Overestimate')
-]
-plt.legend(handles=legend_elements, loc='upper right', frameon=False, fontsize=10)
-
-# Optional: Add counts on bars
-for i, (left, right, match) in enumerate(zip(left_counts, right_counts, match_counts)):
-    if left > 0:
-        plt.text(-left - max_bar*0.05, left_positions[i], str(left), va='center', ha='right', fontsize=9, color='white', fontweight='bold')
-    if right > 0:
-        plt.text(right + max_bar*0.05, right_positions[i], str(right), va='center', ha='left', fontsize=9, color='white', fontweight='bold')
-    if match > 0:
-        plt.text(match_counts[i]/2, positions[i], str(match), va='center', ha='center', fontsize=8, color='black')
-
-plt.tight_layout()
-
-# --- 8. Save & Show ---
-os.makedirs(os.path.dirname(figure_save_path), exist_ok=True)
-plt.savefig(figure_save_path, format='svg', bbox_inches='tight')
-print(f"✅ Diverging bar plot saved to {figure_save_path}")
-
-plt.show()
-
-##
-
-
-
-# %% Difference Gemini easy
-
-
-# --- 1. Process Error Data ---
-system_names = [name.split('.')[1] for name, _ in functional_systems_to_plot]
-plot_list = []
-
-for gt_col, res_col in functional_systems_to_plot:
-    sys_name = gt_col.split('.')[1]
-
-    # Robust parsing
-    gt = df[gt_col].apply(safe_parse)
-    res = df[res_col].apply(safe_parse)
-    error = res - gt
-
-    # Calculate counts
-    matches = (error == 0).sum()
-    under = (error < 0).sum()
-    over = (error > 0).sum()
-    total = error.dropna().count()
-
-    # Calculate Percentages
-    # Using max(total, 1) to avoid division by zero
-    divisor = max(total, 1)
-    match_pct = (matches / divisor) * 100
-    under_pct = (under / divisor) * 100
-    over_pct = (over / divisor) * 100
-
-    plot_list.append({
-        'System': sys_name.replace('_', ' ').title(),
-        'Matches': matches,
-        'MatchPct': match_pct,
-        'Under': under,
-        'UnderPct': under_pct,
-        'Over': over,
-        'OverPct': over_pct
-    })
-
-stats_df = pd.DataFrame(plot_list)
-
-# --- 2. Plotting ---
-fig, ax = plt.subplots(figsize=(12, 8)) # Slightly taller for multi-line labels
-
-color_under = '#E74C3C'
-color_over = '#3498DB'
-bar_height = 0.6
-
-y_pos = np.arange(len(stats_df))
-
-ax.barh(y_pos, -stats_df['Under'], bar_height, label='Under-scored', color=color_under, edgecolor='white', alpha=0.8)
-ax.barh(y_pos, stats_df['Over'], bar_height, label='Over-scored', color=color_over, edgecolor='white', alpha=0.8)
-
-# --- 3. Aesthetics & Labels ---
-
-for i, row in stats_df.iterrows():
-    # Constructing a detailed label for the left side
-    # Matches (Bold) | Under % | Over %
-    label_text = (
-        f"$\mathbf{{{row['System']}}}$\n"
-        f"Matches: {int(row['Matches'])} ({row['MatchPct']:.1f}%)\n"
-        f"Under: {int(row['Under'])} ({row['UnderPct']:.1f}%) | Over: {int(row['Over'])} ({row['OverPct']:.1f}%)"
-    )
-
-    # Position text to the left of the x=0 line
-    ax.text(ax.get_xlim()[0] - 0.5, i, label_text, va='center', ha='right', fontsize=9, color='#333333', linespacing=1.3)
-
-# Zero line
-ax.axvline(0, color='black', linewidth=1.2, alpha=0.7)
-
-# Clean up axes
-ax.set_yticks([])
-ax.set_xlabel('Number of Patients with Error', fontsize=11, fontweight='bold', labelpad=10)
-#ax.set_title('Directional Error Analysis by Functional System', fontsize=14, pad=30)
-
-# Make X-axis labels absolute
-ax.set_xticklabels([int(abs(tick)) for tick in ax.get_xticks()])
-
-# Remove spines
-for spine in ['top', 'right', 'left']:
-    ax.spines[spine].set_visible(False)
-
-# Legend
-ax.legend(loc='upper right', frameon=False, bbox_to_anchor=(1, 1.1))
-
-# Grid
-ax.xaxis.grid(True, linestyle='--', alpha=0.3)
-
-plt.tight_layout()
-plt.show()
-##
-
-
-
-# %% name
-import pandas as pd
-import matplotlib.pyplot as plt
-import os
-import numpy as np
-
-# --- Configuration & Theme ---
-plt.rcParams['font.family'] = 'Arial'
-figure_save_path = 'project/visuals/functional_systems_magnitude_focus.svg'
-
-# --- 1. Process Error Data with Magnitude Breakdown ---
-system_names = [name.split('.')[1] for name, _ in functional_systems_to_plot]
-plot_list = []
-
-for gt_col, res_col in functional_systems_to_plot:
-    sys_name = gt_col.split('.')[1]
-
-    # Robust parsing
-    gt = df[gt_col].apply(safe_parse)
-    res = df[res_col].apply(safe_parse)
-    error = res - gt
-
-    # Granular Counts
-    matches = (error == 0).sum()
-    u_1 = (error == -1).sum()
-    u_2plus = (error <= -2).sum()
-    o_1 = (error == 1).sum()
-    o_2plus = (error >= 2).sum()
-
-    total = error.dropna().count()
-    divisor = max(total, 1)
-
-    plot_list.append({
-        'System': sys_name.replace('_', ' ').title(),
-        'Matches': matches, 'MatchPct': (matches / divisor) * 100,
-        'U1': u_1, 'U2': u_2plus, 'UnderTotal': u_1 + u_2plus,
-        'UnderPct': ((u_1 + u_2plus) / divisor) * 100,
-        'O1': o_1, 'O2': o_2plus, 'OverTotal': o_1 + o_2plus,
-        'OverPct': ((o_1 + o_2plus) / divisor) * 100
-    })
-
-stats_df = pd.DataFrame(plot_list)
-
-# --- 2. Plotting ---
-fig, ax = plt.subplots(figsize=(13, 8))
-
-# Define Magnitude Colors
-c_under_dark, c_under_light = '#C0392B', '#E74C3C' # Dark Red (-2+), Soft Red (-1)
-c_over_dark, c_over_light   = '#2980B9', '#3498DB' # Dark Blue (+2+), Soft Blue (+1)
-bar_height = 0.6
-y_pos = np.arange(len(stats_df))
-
-# Plot Under-scored (Stacked: -2+ then -1)
-ax.barh(y_pos, -stats_df['U2'], bar_height, color=c_under_dark, label='Under -2+', edgecolor='white')
-ax.barh(y_pos, -stats_df['U1'], bar_height, left=-stats_df['U2'], color=c_under_light, label='Under -1', edgecolor='white')
-
-# Plot Over-scored (Stacked: +1 then +2+)
-ax.barh(y_pos, stats_df['O1'], bar_height, color=c_over_light, label='Over +1', edgecolor='white')
-ax.barh(y_pos, stats_df['O2'], bar_height, left=stats_df['O1'], color=c_over_dark, label='Over +2+', edgecolor='white')
-
-# --- 3. Aesthetics & Table Labels ---
-for i, row in stats_df.iterrows():
-    label_text = (
-        f"$\\mathbf{{{row['System']}}}$\n"
-        f"Match: {int(row['Matches'])} ({row['MatchPct']:.1f}%)\n"
-        f"Under: {int(row['UnderTotal'])} ({row['UnderPct']:.1f}%) | Over: {int(row['OverTotal'])} ({row['OverPct']:.1f}%)"
-    )
-    # Position table text to the left
-    ax.text(ax.get_xlim()[0] - 0.5, i, label_text, va='center', ha='right', fontsize=9, color='#333333', linespacing=1.4)
-
-# Formatting
-ax.axvline(0, color='black', linewidth=1.2)
-ax.set_yticks([])
-ax.set_xlabel('Number of Patients with Error', fontsize=11, fontweight='bold')
-#ax.set_title('Directional Error Magnitude (Under vs. Over Scoring)', fontsize=14, pad=35)
-
-# Absolute X-axis labels
-ax.set_xticklabels([int(abs(tick)) for tick in ax.get_xticks()])
-
-# Remove spines and add grid
-for spine in ['top', 'right', 'left']: ax.spines[spine].set_visible(False)
-ax.xaxis.grid(True, linestyle='--', alpha=0.3)
-
-# Legend with magnitude info
-ax.legend(loc='upper right', frameon=False, bbox_to_anchor=(1, 1.1), ncol=2)
-
-plt.tight_layout()
-plt.show()
-##
-# %% test
-# Diagnose: what are the actual differences?
-print("\n🔍 Raw differences (first 5 rows per system):")
-for gt_col, res_col in functional_systems_to_plot:
-    gt = df[gt_col].apply(safe_parse)
-    res = df[res_col].apply(safe_parse)
-    diff = res - gt
-    non_zero = (diff != 0).sum()
-    # Check if it's due to floating point noise
-    abs_diff = diff.abs()
-    tiny = (abs_diff > 0) & (abs_diff < 1e-10)
-    print(f"{gt_col.split('.')[1]:25s}: non-zero = {non_zero:3d}, tiny = {tiny.sum():3d}, max abs diff = {abs_diff.max():.12f}")
-
-##

figure1.py

@@ -263,60 +263,3 @@ plt.legend(frameon=False, loc='upper center', bbox_to_anchor=(0.5, -0.05))
 plt.tight_layout()
 plt.show()
 ##
-
-
-
-
-# %% name
-import matplotlib.pyplot as plt
-
-# Data
-data = {
-    'Visit': [9, 8, 7, 6, 5, 4, 3, 2, 1],
-    'patient_count': [2, 3, 3, 6, 13, 17, 28, 24, 32]
-}
-
-# Create figure and axis
-fig, ax = plt.subplots(figsize=(10, 6))
-
-# Plot the bar chart
-bars = ax.bar(data['Visit'], data['patient_count'], color='darkblue', label='Patients by Visit Count')
-
-# Add labels and title
-ax.set_xlabel('Visit Number (from last to first)', fontsize=12)
-ax.set_ylabel('Number of Patients', fontsize=12)
-ax.set_title('Patient Visits by Visit Number', fontsize=14)
-
-# Invert x-axis to show Visit 9 on the left (descending order) if desired, but keep natural order (1–9 left to right)
-# For descending order (9→1 from left to right), we'd need to reverse:
-# Visit = data['Visit'][::-1], patient_count = data['patient_count'][::-1]
-# But standard practice is ascending (1 to 9), so we'll sort accordingly:
-# Let's sort by Visit to ensure left-to-right: 1,2,...,9
-
-# Actually, your current Visit list is [9,8,...,1], which is descending.
-# Let's sort by Visit for intuitive left-to-right increasing order:
-sorted_indices = sorted(range(len(data['Visit'])), key=lambda i: data['Visit'][i])
-visit_sorted = [data['Visit'][i] for i in sorted_indices]
-count_sorted = [data['patient_count'][i] for i in sorted_indices]
-
-# Re-plot with sorted x-axis:
-ax.clear()
-bars = ax.bar(visit_sorted, count_sorted, color='darkblue', label='Patients by Visit Count')
-
-# Re-apply labels, etc.
-ax.set_xlabel('Number of Visits', fontsize=12)
-ax.set_ylabel('Number of Unique Patients', fontsize=12)
-#ax.set_title('Number of Patients by Visit Number', fontsize=14)
-
-# Add legend
-ax.legend()
-
-# Improve layout and grid
-ax.grid(axis='y', linestyle='--', alpha=0.7)
-plt.xticks(visit_sorted)  # Ensure all integer visit numbers are shown
-
-# Show the plot
-plt.tight_layout()
-plt.show()
-
-##