Making charts

reschola and charts

This document assumes two things:

• charts are made in ggplot2
• you have a fair understanding of ggplot2, so this is not a guide to using that package.

It also does not reiterate the principles of data visualisation, though it does point to some resources in that vein.

The reschola theme

p <- ggplot(mpg) +
 geom_bar(aes(y = class)) +
 labs(title = "Lots of cars", subtitle = "Count of numbers")
p

p +
 theme_schola("x") # only setting the first parameter to get the right gridlines

p +
 theme_schola("x") +
 scale_x_continuous(expand = flush_axis)

ggplot(mpg) +
 geom_point(aes(cty, hwy)) +
 theme_schola("scatter") +
 labs(title = "Lots of cars", subtitle = "Point by point")

ggplot(mpg) +
 geom_point(aes(cty, hwy), alpha = .2) +
 theme_schola("scatter", base_size = 9, margin_side = 0) +
 labs(title = "Lots of cars", subtitle = "Point by point")

ggplot(mpg) +
 geom_point(aes(cty, hwy), alpha = .2) +
 theme_schola("scatter", base_size = 12, margin_side = 0) +
 labs(title = "Lots of cars", subtitle = "Point by point") +
 theme(panel.background = element_rect(fill = "lightpink"))

p +
 theme_schola("x", multiplot = T) +
 scale_x_continuous(expand = flush_axis) +
 facet_wrap(~ manufacturer, nrow = 2)

p +
 theme_schola("x", multiplot = F) +
 scale_x_continuous(expand = flush_axis) +
 facet_wrap(~ manufacturer, nrow = 2)

ggplot(mpg) +
 geom_point(aes(cty, hwy), alpha = .2) +
 theme_schola("scatter", multiplot = T) +
 labs(title = "Lots of cars", subtitle = "Point by point") +
 facet_wrap(~ class, nrow = 2)

p +
 theme_schola("x", multiplot = T) +
 scale_x_continuous(expand = flush_axis) +
 facet_grid(cyl ~ drv)

ggplot(mpg) +
 geom_point(aes(cty, hwy), alpha = .2) +
 theme_schola("scatter", multiplot = T) +
 labs(title = "Lots of cars", subtitle = "Point by point") +
 facet_grid(cyl ~ class)

ggplot(mpg) +
 geom_point(aes(cty, hwy), alpha = .2) +
 theme_schola("scatter", multiplot = F) +
 labs(title = "Lots of cars", subtitle = "Point by point") +
 facet_grid(cyl ~ class)

Building a diverging Likert plot

[To be completed once the function is in place]

Using colour scales

[To be completed once the scales are in place]

Using custom numeric scales

The chartb below illustrates the custom scales

ggplot(mpg, aes(hwy * 100, cty/max(cty))) +
  geom_point(aes(colour = cty/max(cty), size = hwy*100)) +
  theme_schola() +
  scale_y_percent_cz() +
  scale_x_number_cz() +
  scale_color_binned(labels = label_percent_cz()) +
  scale_size_binned(labels = label_number_cz())

The percent formatter and number formatter also handle the decimal mark correctly. The parameters which you would normally find inside the scales::format_*() functions are accessible directly in the scale_[xy]_continuous() functions.

ggplot(mpg, aes(hwy / 100, cty/max(cty)/100)) +
  geom_point(aes(colour = cty/max(cty), size = hwy*100)) +
  theme_schola() +
  scale_y_percent_cz(accuracy = .1) +
  scale_x_number_cz(accuracy = .1) +
  scale_color_binned(labels = label_percent_cz(accuracy = .1)) +
  scale_size_binned(labels = label_number_cz(accuracy = 0.1))

An English-locale version would look like this, using functions in scales or reexported from hrbrthemes:

ggplot(mpg, aes(hwy * 100, cty/max(cty))) +
  geom_point(aes(colour = cty/max(cty), size = hwy*100)) +
  theme_schola() +
  scale_y_percent(accuracy = .1) +
  scale_x_comma(accuracy = .1) +
  scale_color_binned(labels = scales::label_percent(accuracy = .1)) +
  scale_size_binned(labels = scales::label_number(accuracy = .1))

Sizing images

The schola_[word|pdf] formats have opinionated sizing defaults to ensure that resolution, size and font size in the bitmap (PNG) image are in line with the rest of the document. This is particularly tricky in the context of Word documents, hence the baked in defaults. For this reason, the steps to change images sizes are a bit special - see below.

Here is what the defaults are:

• width is full-page, i.e. 15.98 cm (so the image is sized exactly to fit inside margings without any resizing which would degrade quality)
• resolution at 300 DPI and pixel size to result in sharp image with exactly same font size as document
• height is width divided by the golden ratio, i.e. default height is 9.92 cm

Here is how to manipulate image sizes without messing up anything else, in Word:

• to change the height, set the fig.asp chunk option to more or less than 0.62
• to change the width, set chunk options fig.width and out.width to the same size in centimeters, e.g. for half-width you would do fig.width=7.98/2.54, out.width="7.98cm" (because fig.width is set in inches). Note the quotes around the out.width value - this is a special thing for pandoc to ensure the sizing is reflected in the layout of the Word document. The height in this case will be 7.98 * 0.62 cm (see above)
• so to change both width and height, set the fig.asp param as per above.

For HTML output, you can use e.g. out.width="50%".

For PDF output, you can just set fig.width in inches and the height using fig.asp as above, as long as you use the pdf or cairo_pdf device (using the dev option in the chunk or at per-document level.) Note that the font handling might be a bit tricky for PDF depending on your system.

Image sizing works quite differently between output formats (HTML, PDF, Word). For that reason it is probably the one thing that will look odd or break if you try to render a document written in a reschola template into HTML or PDF without fiddling with sizing, graphic devices or font handling.

Making fonts work

The schola_theme() is set to use the Ubuntu and Ubuntu Condensed font families by default. These may not be present on your system. You can install them with reschola::install_reschola_fonts() and by running reschola::register_reschola_fonts() (only needed on Windows if you use Windows bitmap devices – which are default). On package load, these will be registered with R. If the fonts do not show up in your charts, run extrafont::loadfonts() (once per session). You can also set the option reschola.loadfonts to have reschola register these fonts with the PDF/PostScript ecosystem upon package load. (The fonts may or may not work out of the box when you render into PDF depending on your system – if not, first try using the cairo_pdf device and/or running extrafont::embed_fonts() on the whole resulting PDF document.)

Using fonts

If you end up using different fonts, it is good to choose something that (a) has fixed-width figures in the default set of characters, and (b) has sensible widths or a narrow-width member of the type family (usually called “Condensed”, “Narrow” or some such.)

The hrbrthemes package has a good range of fonts, some of which tick these criteria. Arial Narrow is narrow and has fixed-width figures. IBM Plex Sans, Titillium Sans and Econ Sans all have fixed width figures but not all have Narrow companions.

Public Sans does not have fixed-width figures.

See the hrbrthemes vignette for an explanation of why these criteria matter

Tricks of the trade for ggplot2

• theme_void() for visuals that don’t need any gridlines, axes, etc., such as schematic maps
• aes(…, group = 1) for making all data points in a geom_line() into one line
• key_glyph parameter to geoms to pick a specific shape for the legend key
• geom_sf_label for intelligently labelling spatial features plotted through geom_sf()
• labs(colour = NULL) for switching off the colour legend, labs(x = NULL) to turn off colour label - quicker than theme(axis.label.x = element_blank()) and guides(colour = "none")

Resources on good practice

Kieran Healy. Data Visualisation: a practical introduction. With R code.

Claus Wilke. Fundamentals of Data Visualisation No R code, very strong on principles.

Beyond that and in the world of paper books, anything by Alberto Cairo is solid.

Handy (or just cool) ggplot2 extensions