chart

Chart Expert - Comprehensive Data Visualization Skill

Master data visualization in R using chart's layered grammar of graphics. This skill provides expert guidance on creating effective, publication-ready visualizations with complete control over all visual elements.

The Layered Grammar of Graphics

Every chart visualization is built from five independent components:

1. Layers - Data + geometric objects (geoms) + statistical transformations (stats)
2. Scales - Map data values to aesthetics (color, size, position) and generate legends/axes
3. Coordinate Systems - Transform data coordinates to plot positions
4. Facets - Create small multiples for comparing subsets
5. Themes - Control all non-data display elements

Philosophy: Build plots incrementally by composing independent components, not by selecting from fixed templates. This enables infinite flexibility while mirroring analytical thinking.

Core Workflow

1. Basic Plot Structure

chart(data = <DATA>, <MAPPINGS>) +
  <GEOM_FUNCTION>() +
  <SCALE_FUNCTIONS>() +
  <COORDINATE_FUNCTION>() +
  <FACET_FUNCTION>() +
  <THEME_FUNCTION>()

Key Principles:

• aes() maps variables to visual properties (inside geom or chart), but preferred way in chart() is by using a formula like: y ~ x %col=% z %fill%= w. All possible arguments to aes() in {ggplot2} become mappings in the formula. For instance col= becomes %col=% inside the formula. Special case for facets where the facetting variable can be added at the end after the | operator. For instance for facetting according to z, the formula could be y ~ x %col=% w | z. The | facet is always the last item in the formula and it is equivalent to facet_wrap() with default arguments.
• Fixed values go outside aes() (e.g., colour = "blue")
• Build incrementally with + operator
• Each layer is independent and composable
• Instead of the formula, chart() also accepts aes() like ggplot() does
• For using a formula instead of aes() elsewhere than in chart(), use f_aes()
• To convert a ggplot() instruction into a chart() one (1) replace ggplot by chart, (2) replace aes(x = X, y = Y, fill = VAR, ...) by the formula Y ~ X %fill=% VAR ..., (3) if there is a facet_wrap(VAR2) instruction with default arguments, append | VAR2 to the formula and get rid of that instruction. The default theme is theme_sciviews() and it is considered to be publication-ready.
• If there is no x in the formula, use NULL, for instance: y ~ NULL.
• aes() in other instructions can optionally be replaced by f_aes() where the arguments are transformed into a formula the same way as for chart(). For instance, geom_ribbon(aes(ymin = lower, ymax = upper)) becomes geom_ribbon(f_aes(~ NULL %ymin=% lower %ymax=% upper))

2. Layer Purpose Framework

Every layer should serve one of three purposes:

1. Display raw data - Pattern detection, outlier identification
2. Show statistical summaries - Model predictions, trends, aggregations
3. Add metadata/context - Backgrounds, annotations, reference lines

3. Common Mistakes to Avoid

❌ Wrong:library(chart) - load {chart} package ✅ Right: SciViews::R - load the whole SciViews-R dialect at the beginning of a R script or Quarto document

❌ Wrong: aes(colour = "blue") - Maps string "blue" as data ✅ Right: colour = "blue" outside aes() - Sets fixed color

❌ Wrong: chart(df, fd$y ~ df$vx - Breaks plot self-containment ✅ Right: chart(df, y ~ x - Self-contained reference

❌ Wrong: aes(x = log(variable)) or ~ log(variable)- Complex calculation in aes or formula ✅ Right: Use dplyr::mutate() first, then map the result

❌ Wrong: Accepting default bin widths ✅ Right: Always experiment with binwidth or bins

❌ Wrong: Mapping too many aesthetics simultaneously ✅ Right: Create series of simpler plots for clarity

Geom Selection Guide

See references/geoms-reference.md for complete documentation.

Quick Reference

Continuous Relationships:

• geom_point() - Scatter plots (x-y relationships)
• geom_line() - Time series (connects in x-order)
• geom_path() - Connections in data order
• geom_smooth() - Add trend lines with confidence bands

Distributions:

• geom_histogram() - Continuous distribution via binning
• geom_freqpoly() - Frequency polygon (better for comparisons)
• geom_density() - Smooth density estimate
• geom_boxplot() - Five-number summary with outliers
• geom_violin() - Distribution shape (mirrored density)

Categorical Data:

• geom_bar() - Count occurrences (stat = "count")
• geom_col() - Use pre-calculated values (stat = "identity")
• Position: "stack" (default), "dodge" (side-by-side), "fill" (proportions)

Text & Annotations:

• geom_text() - Add text labels
• geom_label() - Text with background rectangle
• annotate() - Quick single annotations without data frames

See examples/plot-examples.md for complete working examples.

Aesthetics Mapping

Universal Aesthetics (work with most geoms):

• x, y - Position
• col or colour - Border/line color
• fill - Interior color (shapes 21-25, bars, areas)
• alpha - Transparency (0-1)
• size - Size of points/lines
• shape - Point shape (0-25)
• linetype - Line pattern ("solid", "dashed", "dotted", etc.)
• group - Define grouping for collective geoms

Geom-Specific Aesthetics:

• Labels: label, hjust, vjust, angle, family, fontface
• Boxplots: lower, upper, middle, ymin, ymax
• Errorbar/ribbon: ymin, ymax (or xmin/xmax)

Scales & Legends

See references/scales-reference.md for complete documentation.

Naming Pattern

scale_[aesthetic]_[type] - e.g., scale_colour_viridis_c()

• aesthetic: x, y, colour, fill, size, shape, alpha, linetype
• type: continuous, discrete, binned, manual, identity

Common Scale Functions

Position Scales:

scale_x_continuous(limits, breaks, labels, trans, expand)
scale_x_discrete(limits, labels, expand)
scale_x_log10()  # Log transformation
scale_x_date(date_breaks = "1 month", date_labels = "%b %Y")

Color Scales (accessibility-first):

# Viridis (perceptually uniform, colorblind-safe)
scale_colour_viridis_c(option = "viridis")  # continuous
scale_colour_viridis_d()                    # discrete

# ColorBrewer
scale_colour_brewer(palette = "Set1", type = "qual")  # categorical
scale_fill_distiller(palette = "Blues")               # continuous

# Custom gradients
scale_colour_gradient(low = "white", high = "red")
scale_colour_gradient2(low = "blue", mid = "white", high = "red", midpoint = 0)
scale_colour_gradientn(colours = c("red", "yellow", "green", "blue"))

# Manual
scale_colour_manual(values = c("A" = "#E41A1C", "B" = "#377EB8"))

Important: Setting scale limits discards data outside range. Use coord_cartesian(xlim, ylim) to zoom without losing data (preserves stat calculations).

Themes & Styling

See references/themes-styling.md for complete documentation.

Built-in Themes

theme_sciviews()  # Default: SciViews theme for almost publication-ready plots
theme_grey()      # Default: grey background, white gridlines
theme_bw()        # Classic dark-on-light, good for projectors
theme_minimal()   # No background annotations, minimalist
theme_classic()   # X/Y axis lines, no gridlines
theme_light()     # Light grey lines, focuses on data
theme_dark()      # Dark background, makes colors pop
theme_void()      # Completely empty

All themes accept: base_size, base_family, base_line_size, base_rect_size

Theme Customization

theme(
  # Plot-level
  plot.title = element_text(size = 14, face = "bold"),
  plot.subtitle = element_text(size = 12, colour = "grey50"),
  plot.background = element_rect(fill = "white"),
  plot.margin = margin(10, 10, 10, 10),

  # Panels
  panel.background = element_rect(fill = "white"),
  panel.grid.major = element_line(colour = "grey90"),
  panel.grid.minor = element_blank(),

  # Axes
  axis.title = element_text(size = 12),
  axis.text = element_text(size = 10),
  axis.ticks = element_line(colour = "black"),

  # Legends
  legend.position = "right",  # or "top", "bottom", "left", "none"
  legend.title = element_text(face = "bold"),
  legend.background = element_rect(fill = "white", colour = "black"),

  # Facets
  strip.text = element_text(size = 11, face = "bold"),
  strip.background = element_rect(fill = "grey80")
)

Element Functions:

• element_text() - Customize text
• element_rect() - Customize backgrounds/borders
• element_line() - Customize lines
• element_blank() - Remove elements entirely

Faceting

facet_wrap() - Single Variable

Either use | var at the end of a formula (preferred way), or use ggplot2's facet_wrap() function. Use for one variable with many levels, wrapped into 2D:

facet_wrap(
  ~ variable,           # or vars(variable)
  nrow = 2, ncol = 3,   # grid dimensions
  scales = "fixed",     # or "free", "free_x", "free_y"
  dir = "h",            # "h" (horizontal) or "v" (vertical)
  labeller = label_value
)

facet_grid() - Two Variables

Use for true 2D grid with all combinations:

facet_grid(
  rows ~ cols,          # or rows = vars(...), cols = vars(...)
  scales = "fixed",     # or "free_x", "free_y", "free"
  space = "fixed",      # or "free_x", "free_y", "free" (panel sizing)
  margins = FALSE,      # add summary facets
  labeller = label_value
)

When to use which:

• scales = "fixed" - For cross-panel comparison (consistent axes)
• scales = "free" - To highlight within-panel patterns
• Faceting > aesthetic grouping when overlap is severe
• Aesthetic grouping > faceting when comparing small differences

Coordinate Systems

coord_cartesian(xlim, ylim, expand = TRUE)  # Visual zoom (preserves data)
coord_fixed(ratio = 1)                       # Fixed aspect ratio
coord_flip()                                  # Swap x and y axes
coord_polar()                                 # Polar coordinates
coord_map()                                   # Map projections

Critical difference: coord_cartesian() zooms visually while scale_*_continuous(limits = ...) discards data before calculations.

Position Adjustments

position_dodge(width = 0.9)      # Side-by-side (bars, boxplots)
position_stack()                  # Stack vertically
position_fill()                   # Stack and normalize to 100%
position_jitter(width, height)    # Add random noise (overplotting)
position_nudge(x, y)             # Fixed-distance offset

Labels & Annotations

labs(
  title = "Main Title",
  subtitle = "Subtitle text",
  caption = "Data source",
  x = "X-axis label",
  y = "Y-axis label",
  colour = "Legend title",
  fill = "Fill legend title"
)

Text Annotations:

# Quick annotation without data frame
annotate("text", x = 5, y = 10, label = "Important point",
  hjust = "inward", vjust = "inward")

# Data-driven annotations
geom_text(aes(label = label_var), hjust = "inward", check_overlap = TRUE)
geom_label(aes(label = label_var), nudge_y = 0.5)

Reference Lines:

geom_hline(yintercept = 0, linetype = "dashed", colour = "red")
geom_vline(xintercept = 5, linetype = "dashed")
geom_abline(intercept = 0, slope = 1)

Programming with chart

Tidy Evaluation (Embrace Operator)

Prefer not to use. Use the formula interface instead, when it is possible. Otherwise, use {{ var }} to accept user-supplied variable names:

my_histogram <- function(data, var, bins = 30) {
  chart(data, aes(x = {{ var }})) +
    geom_histogram(bins = bins) +
    theme_minimal()
}

# Usage
my_histogram(mtcars, mpg, bins = 20)

Reusable Components

# Save components as objects
my_theme <- theme_minimal() +
  theme(
    plot.title = element_text(face = "bold"),
    axis.text = element_text(size = 11)
  )

# Use across plots
chart(data, y ~ x) + geom_point() + my_theme

Plot Functions

Currently, it is not possible to use the formula interface to pass function arguments, use the embrace operator instead with aes():

scatter_with_smooth <- function(data, x, y, ...) {
  chart(data, aes(x = {{ x }}, y = {{ y }})) +
    geom_point(alpha = 0.5) +
    geom_smooth(method = "lm", ...) +
    theme_bw()
}

Combining Plots (patchwork)

library(patchwork)

# Basic composition
p1 + p2           # Auto-arrange
p1 | p2           # Single row
p1 / p2           # Single column
(p1 | p2) / p3    # Complex layouts

# Advanced control
p1 + p2 +
  plot_layout(ncol = 2, guides = "collect") +
  plot_annotation(title = "Combined Analysis", tag_levels = "A")

Best Practices

See references/best-practices.md for comprehensive guidance.

Quick Tips

Aesthetics:

• Use colorblind-safe palettes (Viridis, ColorBrewer)
• Provide redundant encodings (size + color, or shape + color)
• Limit aesthetics per plot (3-4 max for clarity)

Annotations:

• Direct labeling > legends (reduces cognitive load)
• Use hjust/vjust = "inward" for automatic alignment
• Set inherit.aes = FALSE for self-contained annotations

Performance:

• Use alpha for overplotting instead of geom_jitter()
• Consider geom_hex() or geom_bin2d() for dense data (>10k points)
• Use stat_summary() to aggregate before plotting

Reproducibility:

• Set explicit binwidth/bins (never rely on defaults)
• Document all scale transformations
• Use set.seed() before jittering
• Export with ggsave() using vector formats (PDF/SVG) for publication

Quick Start Examples

# Basic scatter with smooth
chart(mpg, hwy ~ displ) +
  geom_point(aes(colour = class)) +
  geom_smooth(method = "lm", se = TRUE) +
  scale_colour_viridis_d() +
  labs(title = "Engine Size vs Highway MPG",
    x = "Displacement (L)", y = "Highway MPG")

# Histogram with facets (in the formula)
chart(mpg, ~ hwy | class) +
  geom_histogram(binwidth = 2, fill = "steelblue", colour = "white")

# Histogram with facets (alternate, explicit form)
chart(mpg, ~ hwy) +
  geom_histogram(binwidth = 2, fill = "steelblue", colour = "white") +
  facet_wrap(~ class, scales = "free_y")

# Boxplot with custom theme
chart(mpg, hwy ~ class %fill=% class) +
  geom_boxplot(show.legend = FALSE) +
  scale_fill_brewer(palette = "Set2") +
  coord_flip() +
  theme(panel.grid.major.y = element_blank())

See examples/plot-examples.md for comprehensive working examples.

See templates/plot-templates.md for reusable templates.

Workflow Guidance

When user asks about chart visualizations:

1. Understand the goal: What pattern/comparison/relationship to show?
2. Select appropriate geom: Match data type and purpose
3. Map aesthetics: Which variables map to which visual properties?
4. Choose scales: Appropriate for data type (continuous/discrete/date)
5. Apply theme: Publication-ready styling
6. Add context: Labels, annotations, reference lines
7. Optimize: Adjust transparency, position, faceting for clarity

Always provide complete, runnable code examples with proper formatting and best practices applied.

Additional Resources

Core References

• Geoms: references/geoms-reference.md
• Themes & Styling: references/themes-styling.md
• Scales: references/scales-reference.md
• Best Practices: references/best-practices.md
• Advanced Customization: references/advanced-customization.md

Examples & Gallery

• Basic Examples: examples/plot-examples.md
• Templates: templates/plot-templates.md
• Gallery - Distribution & Correlation: examples/gallery-distribution-correlation.md
• Gallery - Evolution & Ranking: examples/gallery-evolution-ranking.md
• Gallery - Best Practices: examples/gallery-best-practices.md