feat(seaborn): implement scatter-lag

[github-actions]

Implementation: scatter-lag - seaborn

Implements the seaborn version of scatter-lag.

File: plots/scatter-lag/implementations/seaborn.py

Parent Issue: #5251

---

🤖 impl-generate workflow

[claude]

AI Review - Attempt 1/3

Image Description

The plot shows a scatter lag plot for an AR(1) time series, displaying y(t) on the x-axis vs y(t+1) on the y-axis. Points are colored using the viridis colormap (purple/dark blue for early time indices, transitioning to green and yellow for later ones), creating a temporal gradient. A dashed diagonal reference line in salmon/reddish color runs from lower-left to upper-right, representing the y=x (zero autocorrelation) baseline. A colorbar on the right labeled "Time Index" spans 0 to ~425. In the upper-left corner, a rounded white annotation box shows "r = 0.83" in bold. The title reads "AR(1) Autocorrelation · scatter-lag · seaborn · pyplots.ai". Both top and right spines are removed, and a very subtle grid is present. The overall color scheme is clean and professional. The scatter cloud clusters tightly along the diagonal, clearly demonstrating strong positive autocorrelation (r=0.83).

Score: 83/100

Category	Score	Max
Visual Quality	27	30
Design Excellence	13	20
Spec Compliance	14	15
Data Quality	15	15
Code Quality	10	10
Library Mastery	4	10
Total	83	100

Visual Quality (27/30)

• VQ-01: Text Legibility (8/8) - All font sizes explicitly set: title 24pt, labels 20pt, ticks 16pt, colorbar label 18pt
• VQ-02: No Overlap (6/6) - No overlapping text elements
• VQ-03: Element Visibility (4/6) - s=70, alpha=0.65 is too large/opaque for 499 points (guidelines: s=20-50, alpha=0.3-0.5 for 300+); pattern is visible but central region is dense
• VQ-04: Color Accessibility (4/4) - viridis is perceptually uniform and colorblind-safe
• VQ-05: Layout & Canvas (4/4) - Good proportions, colorbar well-positioned, nothing cut off
• VQ-06: Axis Labels & Title (1/2) - Labels "y(t)" and "y(t + 1)" are descriptive but lack units

Design Excellence (13/20)

• DE-01: Aesthetic Sophistication (5/8) - Above defaults: custom annotation box, controlled typography colors (#333333, #444444), intentional red reference line (#c44e52) with alpha. Not FiveThirtyEight-level but clearly above "library defaults"
• DE-02: Visual Refinement (4/6) - sns.despine(), very subtle grid (alpha=0.15, linewidth=0.6), white edge colors on markers. Good refinements but not all details polished
• DE-03: Data Storytelling (4/6) - Time coloring + r annotation + diagonal reference create a narrative about AR(1) autocorrelation. Visual hierarchy through the temporal color gradient guides the viewer

Spec Compliance (14/15)

• SC-01: Plot Type (5/5) - Correct lag plot (y(t) vs y(t+1))
• SC-02: Required Features (4/4) - Diagonal reference line, time-indexed coloring, correlation coefficient annotation, configurable lag variable all present
• SC-03: Data Mapping (3/3) - y(t) on x-axis, y(t+1) on y-axis, all data visible
• SC-04: Title & Legend (2/3) - Contains required components but format is "AR(1) Autocorrelation · scatter-lag · seaborn · pyplots.ai" instead of exact "{spec-id} · {library} · pyplots.ai"

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) - Shows all aspects: autocorrelation pattern, diagonal baseline, temporal structure, quantified correlation
• DQ-02: Realistic Context (5/5) - Synthetic AR(1) process data is explicitly listed in spec as an example domain; scientifically valid and neutral
• DQ-03: Appropriate Scale (4/4) - phi=0.85, noise N(0,1) produces realistic AR(1) values in range ~[-5, 5]

Code Quality (10/10)

• CQ-01: KISS Structure (3/3) - Clean linear flow: imports → data → plot → save, no functions or classes
• CQ-02: Reproducibility (2/2) - np.random.seed(42) set
• CQ-03: Clean Imports (2/2) - All four imports (matplotlib, numpy, pandas, seaborn) are used
• CQ-04: Code Elegance (2/2) - Clean Pythonic code, appropriate complexity for the visualization
• CQ-05: Output & API (1/1) - Saves as plot.png, current seaborn 0.14+ API used

Library Mastery (4/10)

• LM-01: Idiomatic Usage (3/5) - Uses sns.set_theme() and sns.despine() correctly, but the core scatter plot uses ax.scatter() (matplotlib) instead of sns.scatterplot(). The preferred seaborn pattern is sns.scatterplot(data=df, x='y(t)', y='y(t + 1)', hue='Time Index', palette='viridis', ax=ax) for continuous hue mapping
• LM-02: Distinctive Features (1/5) - Only sns.set_theme() and sns.despine() represent seaborn-distinctive usage; the primary visualization is matplotlib. No use of seaborn's high-level statistical plotting strengths

Score Caps Applied

• None applied — DE-01=5 and DE-02=4 are both above the cap threshold

Strengths

• Excellent data quality: AR(1) with phi=0.85 creates a textbook autocorrelation example showing clear diagonal clustering
• Perfect visual quality fundamentals: all font sizes explicitly set, viridis colormap, clean layout
• Strong spec compliance: all required features (diagonal line, temporal coloring, r annotation) implemented correctly
• Clean code structure with reproducibility seed

Weaknesses

• Core scatter plot uses matplotlib ax.scatter() instead of seaborn's sns.scatterplot(); misses seaborn's idiomatic approach for continuous hue coloring
• Marker size (s=70) and opacity (alpha=0.65) too large for 499 data points; creates unnecessary density in center
• No seaborn-distinctive statistical features leveraged (e.g., sns.regplot for trend overlay, or seaborn's native hue handling)

Issues Found

1. LM-01/LM-02 LOW: Uses ax.scatter() instead of sns.scatterplot() — seaborn's scatterplot with hue for continuous data is the idiomatic pattern
   • Fix: Use sns.scatterplot(data=df, x='y(t)', y='y(t + 1)', hue='Time Index', palette='viridis', ax=ax, size=..., alpha=...) and extract colorbar from ax.collections
2. VQ-03 PARTIAL: s=70, alpha=0.65 too large/opaque for 499 points
   • Fix: Reduce to s=30-40, alpha=0.4-0.5 per density guidelines
3. SC-04 MINOR: Title prefix "AR(1) Autocorrelation ·" before spec-id deviates from required format
   • Fix: Use "scatter-lag · seaborn · pyplots.ai" as exact title or confirm descriptive prefix is acceptable

AI Feedback for Next Attempt

Switch core scatter to sns.scatterplot() with hue='Time Index' and palette='viridis' for idiomatic seaborn usage — this unlocks seaborn's native hue handling and makes the library mastery evident. Reduce marker size to s=30 and alpha=0.4 for 499 data points. Consider adding sns.regplot or a trend overlay to leverage seaborn's statistical plotting strengths. Fix title to exact format "scatter-lag · seaborn · pyplots.ai".

Verdict: REJECTED

[github-actions]

🔧 Repair Attempt 1/3

Applied fixes based on AI review feedback.

Status: Repair completed, re-triggering review...

---

🤖 impl-repair

[claude]

AI Review - Attempt 2/3

Image Description

The plot shows a scatter lag plot of a synthetic AR(1) time series. The x-axis is labeled "y(t)" and the y-axis "y(t + 1)", both in clean sans-serif font. Approximately 499 scatter points are rendered with the viridis colormap (purple → teal → green → yellow) encoding the time index from 0 to ~499. A red dashed diagonal reference line (y = x) runs from bottom-left to top-right, visually anchoring the correlation pattern. A colorbar on the right side of the plot is labeled "Time Index" with ticks from 0 to 400. In the top-left corner, a rounded annotation box displays "r = 0.83" in bold. The title reads "scatter-lag · seaborn · pyplots.ai". The plot uses the seaborn "ticks" style with top/right spines removed and a very faint grid. The scatter shows a clear linear pattern along the diagonal, indicating strong positive autocorrelation at lag 1. Overall layout is clean and balanced on a 16:9 canvas.

Score: 89/100

Category	Score	Max
Visual Quality	30	30
Design Excellence	13	20
Spec Compliance	15	15
Data Quality	14	15
Code Quality	10	10
Library Mastery	7	10
Total	89	100

Visual Quality (30/30)

• VQ-01: Text Legibility (8/8) — All font sizes explicitly set: title 24pt, labels 20pt, ticks 16pt, colorbar label 18pt, annotation 18pt; all crisp and readable
• VQ-02: No Overlap (6/6) — No text collisions; annotation box, colorbar, axis labels, and title all occupy distinct regions
• VQ-03: Element Visibility (6/6) — s=35, alpha=0.45 is well-calibrated for 499 points (guideline: s=20–50, alpha=0.3–0.5 for 300+); points clearly visible without over-saturation
• VQ-04: Color Accessibility (4/4) — Viridis is perceptually uniform and colorblind-safe; red dashed line provides clear contrast
• VQ-05: Layout & Canvas (4/4) — Balanced margins, plot fills ~65–70% of canvas width; colorbar is appropriately sized
• VQ-06: Axis Labels & Title (2/2) — "y(t)" / "y(t + 1)" is standard lag plot notation; descriptive and domain-appropriate

Design Excellence (13/20)

• DE-01: Aesthetic Sophistication (5/8) — Above well-configured default: muted reference line color (#c44e52 at 60% alpha), viridis for temporal ordering, styled annotation box with rounded corners and subtle border, muted font colors ([area-basic] plotnine implementation #333–feat(plotly): implement donut-labeled #555). Intentional and cohesive, but stops short of publication-ready polish.
• DE-02: Visual Refinement (4/6) — Top/right spines removed via sns.despine; grid is very subtle (alpha=0.15, linewidth=0.6); custom axis edge color (#888888) and linewidth. Good refinement — clearly above defaults, but not every detail polished (e.g., colorbar tick styling is standard).
• DE-03: Data Storytelling (4/6) — Visual hierarchy works well: viridis coloring reveals that autocorrelation persists uniformly across time (no drift); the r=0.83 annotation gives immediate quantitative context; the reference line anchors interpretation. Viewer grasps the story quickly. Minor gap: the data only demonstrates one scenario (strong positive autocorrelation) — showing contrast against a random or weakly correlated lag would strengthen the diagnostic narrative.

Spec Compliance (15/15)

• SC-01: Plot Type (5/5) — Correct scatter of y(t) vs y(t+k) for lag k=1
• SC-02: Required Features (4/4) — Diagonal reference line ✓, points colored by time index ✓, correlation coefficient annotation ✓, configurable lag variable ✓
• SC-03: Data Mapping (3/3) — x=y(t), y=y(t+1); time index mapped to viridis color; axes show full data range
• SC-04: Title & Legend (3/3) — Title format "scatter-lag · seaborn · pyplots.ai" correct; colorbar serves as legend for time index (no separate legend needed)

Data Quality (14/15)

• DQ-01: Feature Coverage (5/6) — AR(1) with phi=0.85 clearly shows strong positive autocorrelation, which is the primary use case for lag plots. Time coloring adds another dimension. Minor deduction: only one autocorrelation scenario shown — a second panel or comment showing what a random (uncorrelated) series looks like would demonstrate the full diagnostic utility
• DQ-02: Realistic Context (5/5) — Synthetic AR(1) process is explicitly listed in the spec as example data ("daily stock returns, hourly temperature readings, or synthetic AR(1) process data"); neutral and scientifically plausible
• DQ-03: Appropriate Scale (4/4) — phi=0.85 with unit-variance noise yields values in [-4, 6], appropriate for a standardized financial/environmental time series

Code Quality (10/10)

• CQ-01: KISS Structure (3/3) — Clean Imports → Data → Plot → Save; no functions or classes
• CQ-02: Reproducibility (2/2) — np.random.seed(42)
• CQ-03: Clean Imports (2/2) — matplotlib.pyplot, numpy, pandas, seaborn — all used, none superfluous
• CQ-04: Code Elegance (2/2) — Clean, Pythonic; appropriate complexity; no fake UI elements
• CQ-05: Output & API (1/1) — plt.savefig("plot.png", dpi=300, bbox_inches="tight"); current seaborn 0.13 API

Library Mastery (7/10)

• LM-01: Idiomatic Usage (4/5) — sns.set_theme with custom rc params, sns.scatterplot with continuous hue, sns.despine — all idiomatic. Minor deduction: the colorbar must be added via matplotlib's ScalarMappable (legend=False was set), which is the correct workaround but reveals a seaborn limitation
• LM-02: Distinctive Features (3/5) — Seaborn's continuous hue encoding in scatterplot with viridis palette is library-distinctive. sns.set_theme theming and sns.despine are seaborn-specific. However, most of the visual decoration (colorbar, annotation, reference line) is handled through matplotlib directly, limiting seaborn's distinctiveness score.

Score Caps Applied

• None — DE-01 (5) > 2 and DE-02 (4) > 2, so "correct but boring" cap does not apply

Strengths

• Perfect spec compliance: diagonal reference line, time-index coloring, and r-value annotation all implemented exactly as specified
• Excellent visual quality: font sizes explicitly set to correct values, viridis colormap is colorblind-safe and semantically meaningful for temporal ordering
• Clean, reproducible code with appropriate marker size/alpha calibration for the dataset density

Weaknesses

• Design excellence is good but not exceptional — the aesthetic is polished but not publication-ready; a more distinctive visual treatment would push DE-01 from 5 to 6–7
• Data storytelling only demonstrates one scenario — strong positive autocorrelation; contrasting with a second panel or annotation about what weak/no autocorrelation looks like would strengthen the diagnostic narrative

Issues Found

1. DE-01 BELOW THRESHOLD: Design is intentional and clean but doesn't reach "publication-ready" polish
   • Fix: Consider a slightly darker, more refined background treatment (e.g., very light #f8f8f8 fill), adjust colorbar aesthetics (remove colorbar spines or use a thinner colorbar), refine the reference line (slightly thicker or with explicit label in legend)
2. DE-03 GOOD BUT NOT EXCELLENT: Only one autocorrelation scenario; the story could be richer
   • Fix: Consider adding a subtitle (plt.suptitle or ax.set_title with subtitle) explaining what the linear pattern means, or annotating "Strong AR(1) autocorrelation" as context text

AI Feedback for Next Attempt

Improve design excellence to cross the 90-point threshold: (1) Refine the colorbar — reduce its width (shrink=0.6), remove its spines, and label it with slightly more context (e.g., "Observation Index"); (2) Add a meaningful subtitle or secondary annotation that contextualizes the diagnostic result (e.g., text below the r annotation explaining "Linear pattern → strong AR(1) dependence"); (3) Slightly increase the reference line linewidth to 2.5 and reduce its alpha to 0.5 for a cleaner look; (4) Consider a very subtle background fill (ax.set_facecolor('#fafafa')) to add depth. These tweaks should push DE-01 to 6/8 and DE-03 to 5/6, reaching ~91–92 total.

Verdict: REJECTED