Publication Figures

When to Use

• Creating or editing any figure, plot, chart, or multi-panel export
• Generating artifacts under reports/
• Any prompt mentioning figure, plot, panel, chart, axis, legend, DPI

Required Output

• Vector PDF + 600 DPI PNG for every figure (export both).
• Font: Arial (fall back to a sans-serif that renders as Arial); embed fonts in the PDF.
• Palette: Okabe-Ito colorblind-safe palette; verify the figure is legible in grayscale.
• Statistics on the figure: show error bars, the n per group, and significance markers where a comparison is made.
• No chartjunk: minimal gridlines, direct labels over legends where feasible.

Matplotlib Setup Snippet

import matplotlib as mpl
mpl.rcParams.update({
    "font.family": "Arial",
    "pdf.fonttype": 42,   # embed TrueType (editable text in the PDF)
    "ps.fonttype": 42,
    "savefig.dpi": 600,
    "figure.constrained_layout.use": True,
})

# Okabe-Ito
OKABE_ITO = ["#000000", "#E69F00", "#56B4E9", "#009E73",
             "#F0E442", "#0072B2", "#D55E00", "#CC79A7"]

def save_figure(fig, stem):
    fig.savefig(f"reports/{stem}.pdf")           # vector
    fig.savefig(f"reports/{stem}.png", dpi=600)  # raster

Anti-Patterns

• Rainbow / jet colormaps, or palettes indistinguishable in grayscale.
• PNG-only export (no vector), or DPI < 300.
• Bare point estimates with no error bars, n, or significance.

Integration

• Pair with [[data-analysis-hygiene]] — audit the data before plotting it.