feat(matplotlib): implement survival-kaplan-meier

[github-actions]

Implementation: survival-kaplan-meier - matplotlib

Implements the matplotlib version of survival-kaplan-meier.

File: plots/survival-kaplan-meier/implementations/matplotlib.py

---

🤖 impl-generate workflow

[claude]

AI Review - Attempt 1/3

Image Description

The plot displays a Kaplan-Meier survival analysis comparing two groups: Treatment Group (blue, #306998) and Control Group (yellow/gold, #FFD43B). Both curves are step functions starting at survival probability 1.0 and decreasing over time (0-60 months). The Treatment Group shows better survival with a median of 21.9 months vs Control Group median of 16.6 months. Each curve has a semi-transparent 95% confidence interval band. Vertical tick marks along the curves indicate censored observations. Dotted vertical lines mark the median survival times, with a horizontal dotted line at 0.5 probability. A text box in the bottom-right corner displays the median values. The legend is positioned in the upper-right corner. Grid is subtle with alpha=0.3 and dashed lines.

Quality Score: 92/100

Criteria Checklist

Visual Quality (37/40 pts)

• VQ-01: Text Legibility (10/10) - Title at 24pt, axis labels at 20pt, ticks at 16pt, legend at 16pt - all perfectly readable
• VQ-02: No Overlap (8/8) - No overlapping text elements
• VQ-03: Element Visibility (8/8) - Step curves with linewidth=3, tick marks with s=400 are clearly visible
• VQ-04: Color Accessibility (5/5) - Blue and yellow/gold are colorblind-safe, excellent contrast
• VQ-05: Layout Balance (4/5) - Good proportions, plot fills canvas well, slight issue with confidence bands extending to edges
• VQ-06: Axis Labels (2/2) - "Time (months)" and "Survival Probability" are descriptive with units
• VQ-07: Grid & Legend (0/2) - Grid is subtle (alpha=0.3), but legend is in upper right which could potentially overlap with data in some scenarios; also median text box partially overlaps with y-axis area

Spec Compliance (25/25 pts)

• SC-01: Plot Type (8/8) - Correct step function Kaplan-Meier curves
• SC-02: Data Mapping (5/5) - Time on X, survival probability on Y
• SC-03: Required Features (5/5) - Step functions, 95% CI bands, censored tick marks, group comparison, median annotations all present
• SC-04: Data Range (3/3) - X: 0-65 months, Y: 0-1.05 shows all data
• SC-05: Legend Accuracy (2/2) - Treatment Group and Control Group labels correct
• SC-06: Title Format (2/2) - "survival-kaplan-meier · matplotlib · pyplots.ai" matches spec

Data Quality (18/20 pts)

• DQ-01: Feature Coverage (7/8) - Shows two groups with different survival curves, censoring, CI widening over time; could have shown more dramatic early separation
• DQ-02: Realistic Context (7/7) - Clinical trial survival data is a perfect real-world application
• DQ-03: Appropriate Scale (4/5) - 60 months follow-up is realistic, but the exponential scale=24 for treatment vs scale=16 for control could show clearer separation

Code Quality (9/10 pts)

• CQ-01: KISS Structure (3/3) - Linear flow: imports → data → K-M calculation → plot → save
• CQ-02: Reproducibility (3/3) - np.random.seed(42) is set
• CQ-03: Clean Imports (2/2) - Only matplotlib.pyplot and numpy used, both necessary
• CQ-04: No Deprecated API (1/1) - Uses current matplotlib API
• CQ-05: Output Correct (0/1) - Saves as plot.png (correct)

Library Features (3/5 pts)

• LF-01: Uses distinctive library features (3/5) - Good use of ax.step with where="post", fill_between with step="post", scatter for tick marks, but no advanced matplotlib features like spines customization or twin axes for at-risk table

Strengths

• Excellent implementation of Kaplan-Meier step functions using ax.step with correct "post" interpolation
• Complete spec compliance: CI bands, censored tick marks, median annotations, group comparison all included
• Clean, well-commented code following KISS principles with proper Greenwald variance calculation
• Good color choices (blue/yellow) that are colorblind-accessible and high contrast
• Proper handling of censored observations with interpolated survival values for tick mark placement

Weaknesses

• Missing at-risk table below the plot which was suggested in spec Notes
• The confidence interval bands could be slightly more transparent to better show curve overlap regions
• Grid/legend score reduced: median text box could be positioned to avoid any overlap with axis labels

Verdict: APPROVED