import pandas as pd import matplotlib.pyplot as plt import seaborn as sns import os import scipy.stats as stats # Set style for academic charts # Use Times New Roman for publication quality plt.rcParams['font.family'] = 'Times New Roman' # Fallback for Chinese characters if needed, but for English publication Times New Roman is standard # plt.rcParams['font.sans-serif'] = ['SimHei', 'Microsoft YaHei', 'Arial', 'DejaVu Sans'] plt.rcParams['axes.unicode_minus'] = False def reproduce_hawthorne(): # Locate data file possible_paths = [ "hawthorne_plot_data.xlsx", "db/hawthorne_plot_data.xlsx", os.path.join(os.path.dirname(__file__), "hawthorne_plot_data.xlsx"), os.path.join(os.path.dirname(__file__), "db/hawthorne_plot_data.xlsx") ] data_path = None for path in possible_paths: if os.path.exists(path): data_path = path break if not data_path: print("Error: Could not find 'hawthorne_plot_data.xlsx'. Please place it next to this script.") return print(f"Loading data from: {data_path}") df = pd.read_excel(data_path) # Verify Data n = len(df) k = max(1, int(n * 0.25)) # First/Last 25% # --- ADAPTATION FOR NEW RESEARCH DESIGN (Section 5.4.2) --- # The new design focuses on "Total Cognitive Depth Score" rather than just "Active Students". # If the column exists, use it. If not, derive it from 'Total Interaction Score' # as a proxy for demonstration purposes (assuming Score ~ Depth). target_metric = 'Total Cognitive Depth Score' # Calculate Total Cognitive Depth Score based on Total Interaction Score # The user requested "Total Cognitive Depth Score" (深度认知总分数). # Since the score is already weighted (High-level +20, General +10), # the Total Interaction Score itself serves as the Total Cognitive Depth Score. df['Total Cognitive Depth Score'] = df['Total Interaction Score'] # 3. Longitudinal Trend Analysis (Linear Regression) # We want to see if the "Total Cognitive Depth Score" increases over time (Lesson Index) slope, intercept, r_value, p_value, std_err = stats.linregress(df['Lesson Index'], df['Total Cognitive Depth Score']) # --- FIX: SMOOTHING THE SPIKE (Lesson 18) --- # Apply a 3-lesson moving average to smooth out the spike at Lesson 18 and other fluctuations # This makes the trend clearer without hiding data (we can mention smoothing in caption) df['Smoothed Score'] = df['Total Cognitive Depth Score'].rolling(window=3, center=True).mean() # Fill NaN at edges with original values to keep length consistent df['Smoothed Score'] = df['Smoothed Score'].fillna(df['Total Cognitive Depth Score']) # Use smoothed score for plotting, but keep original for stats calculation to be honest? # Actually, user suggested smoothing for visual clarity. Let's use smoothed score for the line plot. # Recalculate averages for Step Change visualization based on RAW data (to be honest) or SMOOTHED? # User said: "In Initial Phase (Lesson 1-6) draw a horizontal line... In Final Phase (Lesson 20-25)..." # Let's use the raw data averages for the "Step Change" lines to be statistically accurate, # but plot the smoothed line for visual trend. # Define phases phase1_end = 6 phase2_start = 20 # Calculate means for Step Change mean_phase1 = df[df['Lesson Index'] <= phase1_end]['Total Cognitive Depth Score'].mean() mean_phase2 = df[df['Lesson Index'] >= phase2_start]['Total Cognitive Depth Score'].mean() growth_rate = ((mean_phase2 - mean_phase1) / mean_phase1) * 100 print(f"Verified Growth Rate ({target_metric}): {growth_rate:.2f}% (Lessons 1-{phase1_end} vs {phase2_start}-{len(df)})") # Create Output Dir output_dir = "reproduced_figures" if not os.path.exists(output_dir): os.makedirs(output_dir) # Plot (English Version) plot_chart(df, mean_phase1, mean_phase2, growth_rate, phase1_end, phase2_start, output_dir, lang='en') # Plot (Chinese Version) plot_chart(df, mean_phase1, mean_phase2, growth_rate, phase1_end, phase2_start, output_dir, lang='cn') print(f"Reproduced charts saved to: {os.path.abspath(output_dir)}") def plot_chart(df, mean1, mean2, growth_rate, p1_end, p2_start, output_dir, lang='en'): fig, ax = plt.subplots(figsize=(12, 6)) # Configuration if lang == 'en': ylabel = "Total Substantive Engagement Score" xlabel = "Lesson Sequence" title = "" phase1_text = "Adaptation Phase" phase2_text = "Stabilisation Phase" # Changed from Intervention Effect Phase as requested # Phase 2 is Intervention Phase (7-19), Phase 3 is Stabilisation Phase (20-25) # But our code currently uses 2 phases for growth calculation. # User requested: Adaptation Phase (1-6) | Intervention Phase (7-19) | Stabilisation Phase (20-25) # We should update the visual phases. filename = "Fig3_Hawthorne_Trend.png" smoothed_label = "3-Lesson Moving Average" raw_label = "Raw Data" else: # For Chinese chart, we might need a Chinese font. # Temporarily switch font for Chinese output plt.rcParams['font.sans-serif'] = ['SimHei', 'Microsoft YaHei', 'Arial'] plt.rcParams['font.family'] = 'sans-serif' ylabel = "深度认知总分数 (分)" xlabel = "课时序列" title = "" phase1_text = "适应期" phase2_text = "稳定期" filename = "Fig3_Hawthorne_Trend_cn.png" smoothed_label = "3课时移动平均" raw_label = "原始数据" # 1. Plot Raw Data (Lightly, to show honesty) sns.lineplot(data=df, x='Lesson Index', y='Total Cognitive Depth Score', ax=ax, color='lightgray', alpha=0.5, linewidth=1, label=raw_label, zorder=1) # 2. Plot Smoothed Data (Bold, to show trend) sns.lineplot(data=df, x='Lesson Index', y='Smoothed Score', ax=ax, color='#2c7bb6', linewidth=2.5, marker='o', markersize=6, label=smoothed_label, zorder=2) # 3. Step Change Visualization (The "Grandpa's Trick") # Horizontal line for Phase 1 ax.hlines(y=mean1, xmin=1, xmax=p1_end, colors='#d7191c', linestyles='-', linewidth=3, zorder=3) # Horizontal line for Phase 3 (Stabilisation Phase, 20-25) - previously called Phase 2 in code ax.hlines(y=mean2, xmin=p2_start, xmax=df['Lesson Index'].max(), colors='#d7191c', linestyles='-', linewidth=3, zorder=3) # Add text for means ax.text(p1_end/2 + 0.5, mean1 + 2, f"Mean = {mean1:.1f}", color='#d7191c', ha='center', fontweight='bold', fontsize=12) ax.text((p2_start + df['Lesson Index'].max())/2, mean2 + 2, f"Mean = {mean2:.1f}", color='#d7191c', ha='center', fontweight='bold', fontsize=12) # Arrow and Growth Rate mid_x = (p1_end + p2_start) / 2 # Position text box much higher, above the highest point in the graph or comfortably between the two mean lines # Option 1: Place it between the two red lines, centered. mid_y_between_lines = (mean1 + mean2) / 2 # Option 2: Place it clearly above the lower line, pointing to the upper line? # User suggestion: "Use a dashed large arrow connecting front and back red lines" # Let's draw a dashed connection line from mean1 to mean2 ax.plot([p1_end, p2_start], [mean1, mean2], color='#d7191c', linestyle=':', linewidth=2, zorder=3) # Place the text box in the middle of this connection line ax.text(mid_x, mid_y_between_lines + 50, f"+{growth_rate:.1f}%", ha='center', va='bottom', fontsize=14, fontweight='bold', color='#d7191c', bbox=dict(boxstyle="round,pad=0.3", fc="white", ec="#d7191c", alpha=0.9)) # Draw arrow from text to upper line # ax.annotate("", xy=(p2_start, mean2), xytext=(mid_x, mid_y_between_lines + 50), # arrowprops=dict(arrowstyle="->", color='#d7191c', lw=1.5)) # Styling if title: ax.set_title(title, fontsize=14, fontweight='bold') # Font size increase (User requested 1-2 sizes larger) ax.set_xlabel(xlabel, fontsize=14, fontweight='bold') # Was 12 ax.set_ylabel(ylabel, fontsize=14, fontweight='bold') # Was 12 # Legend customization (User requested transparent border) legend = ax.legend(loc='upper left', fontsize=12) legend.get_frame().set_alpha(0.0) # Make background transparent legend.get_frame().set_linewidth(0.0) # Remove border # Grid customization (User requested minimal grid) # Only horizontal lines, very light ax.yaxis.grid(True, linestyle='--', alpha=0.3) ax.xaxis.grid(False) # Remove vertical grid lines # Phase Separators ax.axvline(x=p1_end + 0.5, color='gray', linestyle=':', alpha=0.4) ax.axvline(x=p2_start - 0.5, color='gray', linestyle=':', alpha=0.4) # Phase Labels (Black, Clear Font) # Get current y-axis limit to place text appropriately y_min, y_max = ax.get_ylim() # Ensure there is enough headroom for labels if y_max < 65: ax.set_ylim(y_min, 65) y_max = 65 text_y_pos = y_max * 0.92 # Adaptation Phase (1-6) ax.text(p1_end/2 + 0.5, text_y_pos, phase1_text, ha='center', color='black', fontsize=12, fontweight='bold') # Intervention Phase (7-19) - Middle Section middle_phase_text = "Intervention Phase" if lang == 'en' else "干预实施期" middle_x = (p1_end + 0.5 + p2_start - 0.5) / 2 ax.text(middle_x, text_y_pos, middle_phase_text, ha='center', color='black', fontsize=12, fontweight='bold') # Stabilisation Phase (20-25) ax.text((p2_start + df['Lesson Index'].max())/2, text_y_pos, phase2_text, ha='center', color='black', fontsize=12, fontweight='bold') # Remove spines ax.spines['top'].set_visible(False) ax.spines['right'].set_visible(False) # Save output_path = os.path.join(output_dir, filename) plt.tight_layout() plt.savefig(output_path, dpi=300) plt.close() print(f"Generated {filename}") if __name__ == "__main__": reproduce_hawthorne()