feat(letsplot): implement residual-plot

[github-actions]

Implementation: residual-plot - letsplot

Implements the letsplot version of residual-plot.

File: plots/residual-plot/implementations/letsplot.py

---

🤖 impl-generate workflow

[claude]

AI Review - Attempt 1/3

Image Description

The plot displays a residual plot with 150 data points. The main visualization uses coral/red colored scatter points for normal residuals and blue colored points for outliers (beyond ±2 standard deviations). A solid blue horizontal reference line at y=0 represents perfect predictions. Two red dashed horizontal lines mark the ±2σ outlier boundaries at approximately ±26.5. A yellow LOWESS smoothing curve clearly shows a U-shaped pattern, correctly revealing the non-linear relationship in the residuals (due to the deliberate quadratic term in the data). The x-axis shows 'Fitted Values' ranging from ~40 to ~480, and the y-axis shows 'Residuals (Observed - Predicted)' ranging from ~-25 to ~30. The legend on the right side distinguishes 'Outlier (>2σ)' and 'Normal' point types. The title correctly follows the format: 'residual-plot · letsplot · pyplots.ai'. The overall layout is clean with a minimal theme and subtle gray gridlines.

Quality Score: 93/100

Criteria Checklist

Visual Quality (37/40 pts)

• VQ-01: Text Legibility (10/10) - Title is bold at 24pt, axis titles at 20pt, tick labels at 16pt, legend text at 16pt - all perfectly readable
• VQ-02: No Overlap (8/8) - No overlapping text elements, all labels clearly visible
• VQ-03: Element Visibility (7/8) - Points are well-sized (size=5) with appropriate alpha (0.7) for 150 data points; slight deduction as some overlapping points in dense areas could benefit from slightly smaller size
• VQ-04: Color Accessibility (5/5) - Blue (#306998) vs coral/red (#DC2626) provides good contrast and is colorblind-distinguishable
• VQ-05: Layout Balance (5/5) - Plot fills canvas well, balanced margins, legend well-positioned
• VQ-06: Axis Labels (2/2) - Descriptive labels: 'Fitted Values' and 'Residuals (Observed - Predicted)'
• VQ-07: Grid & Legend (0/2) - Grid uses alpha=0.5 which is slightly too prominent (recommended 0.2-0.4), and the panel_grid_major setting makes gridlines more visible than ideal

Spec Compliance (25/25 pts)

• SC-01: Plot Type (8/8) - Correct residual plot type with residuals vs fitted values
• SC-02: Data Mapping (5/5) - Fitted values on x-axis, residuals on y-axis as specified
• SC-03: Required Features (5/5) - Includes: horizontal y=0 reference line, LOWESS smoothing, ±2σ outlier bands, outlier coloring
• SC-04: Data Range (3/3) - All data points visible within axes
• SC-05: Legend Accuracy (2/2) - Legend correctly identifies 'Normal' and 'Outlier (>2σ)' point types
• SC-06: Title Format (2/2) - Uses correct format: 'residual-plot · letsplot · pyplots.ai'

Data Quality (19/20 pts)

• DQ-01: Feature Coverage (8/8) - Shows heteroscedasticity (increasing variance with X), non-linear pattern (U-shaped LOWESS), outliers at both extremes, and residuals scattered around zero
• DQ-02: Realistic Context (6/7) - Plausible regression scenario with deliberate model misspecification; slightly generic as it doesn't name a specific domain
• DQ-03: Appropriate Scale (5/5) - Values are sensible for a regression context

Code Quality (9/10 pts)

• CQ-01: KISS Structure (3/3) - Linear structure: imports → data → plot → save
• CQ-02: Reproducibility (3/3) - Uses np.random.seed(42)
• CQ-03: Clean Imports (1/2) - Uses wildcard import with noqa comment (acceptable for lets-plot but not ideal)
• CQ-04: No Deprecated API (1/1) - Uses current lets-plot API
• CQ-05: Output Correct (1/1) - Saves as plot.png and plot.html

Library Features (3/5 pts)

• LF-01: Uses distinctive library features (3/5) - Good use of ggplot grammar with geom_hline, geom_smooth(method='loess'), theme_minimal(), and proper scaling. Could have used lets-plot's interactive tooltip features for HTML output.

Strengths

• Excellent implementation of all spec requirements including y=0 reference line, LOWESS smoothing, ±2σ outlier bands, and outlier coloring
• Clear U-shaped LOWESS curve effectively demonstrates non-linearity detection in residuals
• Good data generation with deliberate heteroscedasticity and non-linear term to showcase residual analysis
• Clean ggplot2-style grammar with proper lets-plot idioms
• Proper sizing (1600x900 base with scale=3) for 4800x2700 output
• Both PNG and HTML outputs generated correctly

Weaknesses

• Grid alpha at 0.5 is slightly too prominent; recommended 0.2-0.4 for subtler appearance
• Wildcard import pattern, while necessary for lets-plot, could be more explicit
• Could leverage lets-plot's interactive tooltips in HTML output for enhanced exploration

Verdict: APPROVED