Figure Table

Axes and labels

• Label axes with full names and units (e.g., "Forecast horizon (weeks)")
• Avoid abbreviations unless defined in the caption
• Use scale_x_date(date_labels = "%b %Y") for time axes
• Start count/score axes at zero unless strong reason not to
• Log-scale axes: label with original values, not log values

Specific plot types

Coefficient/forest plots:

• Point estimate with CI whiskers
• Reference line at null (0 for additive, 1 for multiplicative)
• Order by effect size or grouping variable
• Show both unadjusted (open symbol) and adjusted (filled symbol) if comparing

Time series:

• Observed data as points or thin line
• Predictions/smooths as thicker line with ribbon for CI
• Vertical lines for key events (variant emergence, policy changes)

Score distributions:

• Box plots or violin plots grouped by model characteristic
• Log-scale y-axis for skewed scores
• Show individual model points overlaid if few models

Table conventions

Formatting with kableExtra

library(kableExtra)

df |>
  kbl(
    caption = "Table 1. Descriptive characteristics of included models.",
    digits = 2,
    col.names = c("Characteristic", "N", "Median (IQR)")
  ) |>
  kable_styling(full_width = FALSE) |>
  pack_rows("Model type", 1, 3) |>
  pack_rows("Geographic scope", 4, 5) |>
  footnote(general = "IQR = interquartile range.")

Content conventions

• Table 1: Descriptive characteristics by exposure group
• Include: N (%), median (IQR) or mean (SD) as appropriate
• Group rows logically with pack_rows()
• All abbreviations explained in footnotes
• Numbers in brackets: explicitly label what they are (SD, 95% CI, IQR)

Captions

Captions must be standalone. A reader should understand the figure or table without reading the main text.

Structure:

1. What is shown (the data or result)
2. How it was measured or computed
3. How to interpret (what do colours, panels, axes represent)
4. Abbreviations defined

Example:

Figure 2. Adjusted partial effects of model method on forecast accuracy. Panel A shows effects on case forecasts; Panel B shows effects on death forecasts. Points indicate posterior mean partial effects from the GAMM; whiskers show 95% confidence intervals. The dashed line indicates no difference from the overall mean. WIS = weighted interval score; GAMM = generalised additive mixed model.

Export for journal submission

ggsave(
  here("output", "plots", "figure-1.tif"),
  plot = fig1,
  width = 170, height = 120, units = "mm",
  dpi = 300
)

• Format: TIFF or EPS for submission; PNG for drafts
• Minimum 300 DPI
• Dimensions in mm matching journal specifications
• Separate files for each figure
• Supplementary figures follow the same quality standards