Introduction to programming for data analysis

Session 3 — Data Visualization

Ulrich Aiounou

dimitri.aiounou@univ-amu.fr

AMSE, Aix Marseille University

Ewen Gallic

ewen.gallic@univ-amu.fr

AMSE, Aix Marseille University

Disclaimer

The slides were adapted from those of Pierre Michel and Morgan Raux, both researchers at AMSE, who kindly shared their work.
This slide deck was made with Quarto and reveal.js, it is translated by Claude Sonnet 5 from a LaTex presentation previously made with beamer.

The 4 steps in any data analysis

Finding data
Cleaning data to combine several data sources
Producing knowledge out from data (summary statistics, regression analysis summarised via tables and graphs)
Interpreting the results (writing)

Example of a graph made with `R`

Monthly number of deaths in France (2020–2025). Source: Insee (ref.: 000436394). (R Codes)

A grammar of graphics

Visualization with `{ggplot2}`

There are multiple packages to produce graphs in R.
- The {base} and {ggplot2} packages offer plot functions I rely on the most.
Here, we will introduce only functions from {ggplot2}.
With this package, plots are created following a “grammar of graphics” and are constructed by adding layers to one another.
The (awesome) official documentation: https://ggplot2.tidyverse.org/reference/

library(ggplot2)

Disclaimer The slides are strongly inspired by the 🌐 introduction provided by de Bruin (2024).

What is a grammar of graphics?

A grammar of a language defines the rules of structuring words and phrases into meaningful expressions.
Similarly, a grammar of graphics defines the rules of structuring mathematical and aesthetic elements into a meaningful graph.
Wilkinson (2005) designed the grammar upon which {ggplot2} is based.
To learn with a book: Wickham (2016).

The main elements in a `ggplot2` graph

Data: variables mapped to aesthetic features of the graph.
Geoms: objects/shapes on the graph.
Stats: statistical transformations that summarize data (e.g. mean, confidence intervals).
Scales: mappings of aesthetic values to data values. Legends and axes visualize scales.
Coordinate systems: the plane on which data are mapped on the graphic.
Faceting: splitting the data into subsets to create multiple variations of the same graph.

The base layer

The base layer of a {ggplot2} graph is created with the ggplot() function.
We need to specify two arguments:
- data: the dataset
- mapping: a set of aesthetic mappings between variables in the data and visual properties
Then, we need to provide at least one layer which describes how to render each observation (usually created with a geom function).

ggplot(
  data = txhousing, 
  mapping = aes(x = volume, y = sales)
) +
  geom_point()

Each layer is added using the + operator.

Our first graph with `{ggplot2}`

p_txhousing <- ggplot(
  data = txhousing,
  mapping = aes(x = volume, y = sales)
) +
  geom_point()

p_txhousing

Aesthetics

Geom layers and overriding aesthetics (1/2)

The layers inherit aesthetics from ggplot().
If provided, new aesthetics will override those inherited from ggplot().

ggplot(
  data = txhousing, 
  mapping = aes(x = volume, y = sales)
) +
  geom_point(mapping = aes(colour = listings))

colour = listings causes the colour of objects to vary with values of the variable listings.

Geom layers and overriding aesthetics (2/2)

ggplot(
  data = txhousing, 
  mapping = aes(
    x = volume, 
    y = sales
  )
) +
  geom_point(
    mapping = aes(colour = listings)
  )

Aesthetic attributes

Multiple attributes are available, depending on the geom used, the most common are:
- x, y: positioning along x-axis and y-axis, resp.,
- colour: colour of objects,
- fill: fill colour of objects (name or hexadecimal code),
- linetype: how lines should be drawn (solid, dashed, dotted, …),
- shape: shape of markers,
- size: area of markers (for points), font size (for text),
- linesize: thickness of the lines,
- alpha: transparency (between 0 (transparent) and 1 (opaque)).

Geoms and aesthetics

Each geom has its required aesthetics.
- E.g., geom_point() requires both x and y.
Some aesthetics are not possible for some geoms.
- E.g., while the shape aesthetic is used by geom_point(), it is not accepted by the geom_bar() function.
Check the help pages or the online documentation, or even the 🎨 gallery.

Usual visualizations

Histogram

ggplot(
  data = txhousing,
  mapping = aes(x = median)
) +
  geom_histogram(colour = "white")

A continuous variable (here: median) is cut into bins (default to 30). The number of observations (count) in each bin is reported.

Distribution of the median housing sales in Texas between 2000 and 2015

Density plot

ggplot(
  data = txhousing,
  mapping = aes(x = median)
) +
  geom_density(colour = "blue")

This type of representation can be viewed as a smoothed version of histograms.

Density of the median housing sales in Texas between 2000 and 2015 (kernel estimation)

Bar plot

ggplot(
  data = diamonds,
  mapping = aes(y = cut)
) +
  geom_bar(
    colour = "blue", fill = "grey"
  )

The width of the bar plot reflects the counts of each y-value.

Number of observation per quality of the cut in the `diamonds` dataset.

Scatter plot

ggplot(
  data = txhousing,
  mapping = aes(
    x = volume, y = sales, 
    colour = listings,
    alpha = median,
    size = inventory
    )
) +
  geom_point() +
  theme(legend.position = "bottom")

Useful to represent covariation between two continuous variables.

Number of housing sales vs. total values of sales in Texas

Line plot

ggplot(
  data = txhousing,
  mapping = aes(
    x = date, y = sales, 
    group = city
    )
) +
  geom_line(
    colour = "blue", alpha = .3
  )

Useful to represent covariation between a continuous variable (y-axis) and an ordered variable (x-axis).

Evolution of the number of housing sales in Texas. Each line represents a city.

Labels

The labs() function allows to change axis, legend, and plot labels.

ggplot(
  data = txhousing, 
  mapping = aes(
    x = volume, y = sales,
    colour = listings
  )) +
  geom_point() +
  labs(
    title = "No. housing sales vs. tot. volume",
    subtitle = "Texas, 2000 – 2015",
    x = "Total value of sales",
    y = "Number of sales",
    colour = "Total active listings"
  )

Exporting Graphs

The `ggsave()` function

The ggsave() function saves the last displayed plot (or a specific ggplot object if provided as argument).
It supports many formats: PNG (good for presentations and web), PDF (for print and publications, i.e., your reports), JPEG, TIFF, SVG, EPS.
Unless specified, in RStudio, the plot size and resolution of the ‘Plots’ tab is used (in inches).

# Assuming p contains a ggplot2 graph:
ggsave(file = "figs/plot.pdf", p, width = 5, height = 3)

Output quality

When exporting a scatterplot, if the number of points is large, producing a PDF might not be a good idea: the file will be huge.
To export a plot in PNG with a good resolution, the number of dots per inch (DPI) can be specified.

# Assuming p contains a ggplot2 graph:
ggsave(
   file = "figs/plot.pdf", p, width = 5, height = 3,
   dpi = 300
)

Scales

Every aesthetic (colour, size, shape, x/y position, etc.) has a corresponding scale function.
Default scales are chosen automatically, but can be customized with a scale_* function.
The type of scale depends on the data type of the variable:
- Continuous variables: gradients, continuous axes, size scales,
- Discrete variables: categorical palettes, shapes, breaks.

Scales with a numeric variable

ggplot(
  data = txhousing,
  mapping = aes(
    x = volume, y = sales
  )
) +
  geom_point(
    mapping = aes(colour = listings)
  ) +
  scale_colour_gradient(
    "Total active listings",
    low = "yellow", high = "red"
  )

Scales with a discrete variable

ggplot(
  data = sample_n(diamonds, size = 1000),
  mapping = aes(
    x = carat, y = price, colour = cut
  )
) +
  geom_point() +
  scale_colour_manual(
    "Cut",
    values = c(
      "Fair" = "black", 
      "Good" = "blue", 
      "Very Good" = "darkgreen", 
      "Premium" = "palegreen", 
      "Ideal" = "yellow"
    )
  )

Facetting

Facetting corresponds to creating small subplots for subsets of data.
Each panel shows the same plot for a different level of a variable.
It makes it easier to compare patterns across groups.
There are two main functions:
- facet_wrap(): 1D wrapping of panels,
- facet_grid(): 2D grid of panels.

Facetting with `facet_wrap()`

ggplot(
  data = sample_n(diamonds, size = 5000),
  mapping = aes(
    x = carat, y = price
  )
) +
  geom_point(size = .5, alpha = .1) +
  facet_wrap(~ cut)

Diamonds prices vs. their weights depending on the quality of the cut.

Facetting with `facet_grid()`

ggplot(
  data = sample_n(
    diamonds, 
    size = 5000
  ),
  mapping = aes(
    x = carat, y = price
  )
) +
  geom_point(size = .5, alpha = .1) +
  facet_grid(clarity ~ cut)

Diamonds prices vs. their weights depending on the quality of the cut and the clarity of the diamond.

To go further

This presentation only showed a glimpse of what ggplot2 has to offer.
To go further: 🌐 https://ggplot2-book.org/

Exercises

Practice with the second tutorial!

Appendix

References

Bruin, J. 2024. “R Graphics: Introduction to Ggplot2.” May 2024. https://stats.oarc.ucla.edu/r/seminars/ggplot2_intro/.

Wickham, Hadley. 2016. Ggplot2: Elegant Graphics for Data Analysis, Second Edition. Springer Cham. https://doi.org/https://doi.org/10.1007/978-3-319-24277-4.

Wilkinson, Leland. 2005. The Grammar of Graphics. Springer New York, NY.

Codes for the graph showing the number of deaths in France (1/3)

# Source: https://www.insee.fr/en/statistiques/serie/000436394
file <- "data/deces_france_insee/monthly_values.csv"
deces <- read_csv2(
    file, skip = 4, 
    col_names = c("period", "deaths", "codes")
)

deces <- 
  deces |> 
  mutate(
    date = ym(period),
    month = month(date, label = TRUE),
    year = year(date)
  )

Codes for the graph showing the number of deaths in France (2/3)

# Prepare data for the plot: highlight 2003 and 2021
data_plot_deaths <- deces |> 
  mutate(
    year_excess = case_when(
      year == 2003 ~ "2003 (Drought)",
      year == 2020 ~ "2020 (Covid-19)",
      TRUE ~ "Other"
    ),
    year_excess = factor(
      year_excess, 
      levels = c("2003 (Drought)", "2020 (Covid-19)", "Other")
    )
  )

Codes for the graph showing the number of deaths in France (3/3)

p_deaths <- ggplot(
  data = data_plot_deaths,
  mapping = aes(x = month, y = deaths, group = year, colour = year_excess)
) +
  geom_line(alpha = .8) +
  scale_colour_manual(
    NULL, values = c(
      "2003 (Drought)" = "blue",  "2020 (Covid-19)" = "red", 
      "Other" = "darkgray"
  )) +
  labs(x = NULL, y = "Deaths") +
  scale_y_continuous(
    labels = scales::label_number(suffix = "", scale = 1, big.mark = ",")
  ) + theme_minimal() + theme(legend.position = "bottom")

⏮ Go back

Introduction to programming for data analysis

Disclaimer

The 4 steps in any data analysis

Example of a graph made with R

A grammar of graphics

Visualization with {ggplot2}

What is a grammar of graphics?

The main elements in a ggplot2 graph

The base layer

Our first graph with {ggplot2}

Aesthetics

Geom layers and overriding aesthetics (1/2)

Geom layers and overriding aesthetics (2/2)

Aesthetic attributes

Geoms and aesthetics

Usual visualizations

Histogram

Density plot

Bar plot

Scatter plot

Line plot

Labels

Labels

Exporting Graphs

The ggsave() function

Output quality

Scales

Scales with a numeric variable

Scales with a discrete variable

Facetting

Facetting

Facetting with facet_wrap()

Facetting with facet_grid()

To go further

Exercises

Appendix

References

Codes for the graph showing the number of deaths in France (1/3)

Codes for the graph showing the number of deaths in France (2/3)

Codes for the graph showing the number of deaths in France (3/3)

Example of a graph made with `R`

Visualization with `{ggplot2}`

The main elements in a `ggplot2` graph

Our first graph with `{ggplot2}`

The `ggsave()` function

Facetting with `facet_wrap()`

Facetting with `facet_grid()`