• Part 1: Why do we use visualizations
• Part 2: How to make good data visualizations
  • A grammar of graphics (Wickham) and ggplot2
  • Perception: Visual channels and type of plots
  • Design: Principles of design
  • Storytelling: Use pre-attentive attributes to guide the reader
• Part 3: Conclusions

Anscombe’s quartet (Anscombe, 1973; Chatterjee & Firat, 2007):

• We will talk about how to decompose this graph into “pieces” (e.g. labels, dots, etc)

• For example we often use barplots to compare quantities, scatterplots to show relationships, and lineplots to track a variable over time.

• Write down four ways the plot in the right has improved the plot in the left

• Compare how you read the figure in the right and in the left

• Every visualization should have one main message (and only one)
• The visualization is designed to communicate that message efficiently
• It helps to write down the message (e.g. on the title of the plot)

## # A tibble: 234 × 11
##    manufacturer model      displ  year   cyl trans drv     cty   hwy fl    class
##    <chr>        <chr>      <dbl> <int> <int> <chr> <chr> <int> <int> <chr> <chr>
##  1 audi         a4           1.8  1999     4 auto… f        18    29 p     comp…
##  2 audi         a4           1.8  1999     4 manu… f        21    29 p     comp…
##  3 audi         a4           2    2008     4 manu… f        20    31 p     comp…
##  4 audi         a4           2    2008     4 auto… f        21    30 p     comp…
##  5 audi         a4           2.8  1999     6 auto… f        16    26 p     comp…
##  6 audi         a4           2.8  1999     6 manu… f        18    26 p     comp…
##  7 audi         a4           3.1  2008     6 auto… f        18    27 p     comp…
##  8 audi         a4 quattro   1.8  1999     4 manu… 4        18    26 p     comp…
##  9 audi         a4 quattro   1.8  1999     4 auto… 4        16    25 p     comp…
## 10 audi         a4 quattro   2    2008     4 manu… 4        20    28 p     comp…
## # ℹ 224 more rows

• displ: engine displacement, in litres
• hwy: highway miles per gallon
• class: “type” of car

plot(x = mpg$displ, y = mpg$hwy)

ggplot(data = mpg, 
       mapping = aes(x = displ, 
                     y = hwy, 
                     color = class)) +
   geom_point()

ggplot(data = mpg, 
       mapping = aes(x = displ, 
                     y = hwy)) +
   geom_point() + facet_grid(~ class + factor(year))

ggplot(data = mpg, 
       mapping = aes(x = displ, 
                     y = hwy, 
                     color = class)) +
   geom_point()

ggplot(data = mpg, mapping = aes(x = displ, y = hwy)) +
   geom_point() +
   facet_wrap(~ class, nrow = 2)

ggplot(data = mpg, mapping = aes(y = class)) +
   geom_bar() 

ggplot(data = mpg, mapping = aes(x = displ, y = class, color = factor(cyl))) +
   geom_point(alpha = 0.5, position = "jitter") + 
  theme_minimal()

Colour, fill, alpha (opacity/transparency)

Linetype, linewidth, size, shape, angle

Position: x, y, xmin, xmax, ymin, ymax, xend, yend

Label, fontface, family

https://ggplot2.tidyverse.org/reference/index.html#section-aesthetics

Lines, bars, bins, boxplots, contours, densities, dotplots, error bars, hexagons, polygons, histograms, jittered points, ranges, maps, paths, points, ribbons, areas, rugmarks, segments, curves, smooths, spokes, labels, text, rasters, rectangles, tiles, violins

https://ggplot2.tidyverse.org/reference/index.html#section-geoms

• Choose a plot (left or right). Write down 1—3 things you like, 1—3 things you would change

Before the break

• Part 1: Why do we visualize data
• Part 2: How to make a good data visualization
  • 2.1 A grammar of graphics

Now

• Part 2: How to make a good data visualization
  • 2.2 Visual channels and type of plots
  • 2.3 Principles of design
  • 2.4 Guiding the reader using pre-attentive attributes
• Part 3: Conclusion

Now we know to make a plot with ggplot(). But how do we know which visual channels (aesthetics) and type of plots (geometry) to use?

Visualization Analysis and Design (Tamara Munzner)

• Message you want to convey: Compact and subcompact cars are very efficient
• What are the most important variables? (displ: engine size; hwy: miles per gallon; class: type of car)

ggplot(data = mpg, 
       mapping = aes(y = class, 
                     x = displ, 
                     color = hwy)) +
   geom_point() + theme_minimal()

ggplot(data = mpg, 
       mapping = aes(y = hwy, 
                     x = displ, 
                     color = class)) +
   geom_point()  + theme_minimal()

Fundamentals of Data Visualization by Claus O. Wilke

Fundamentals of Data Visualization by Claus O. Wilke

Fundamentals of Data Visualization by Claus O. Wilke

Fundamentals of Data Visualization by Claus O. Wilke

Fundamentals of Data Visualization by Claus O. Wilke

• Idea: Unique elements should stand apart from each other
• How: Increase contrast ane eliminate clutter (maximize data-to-ink ratio)

• Repetition creates cohesion
• Make sure the same information is always presented in a coherent way (e.g. same color)

• Proper alignment: every element is visually connected to another elements
• Increases cohesion
• Humans like “strong” lines

• Proximity reduces clutter and organizes the space

• Write down four ways (CRAP) the plot in the right has improved the plot in the left

• Reduce cognitive load:
  • Removing unnecessary clutter
  • More professional/aesthetically pleasant
• Contrast:
  • Eliminate unnecessary lines (all frames, use gray grid lines, etc)
  • Don’t use a gray background
  • White space is your friend (allows for “breathing”)
  • Enlarge the labels
  • Use vector graphics (svg/pdf/eps) to avoid blurry figures –> Edit them in Illustrator or Inkscape
• Repetition: Be consistent across figures
• Alignment: Make sure you align subplots/labels
• Proximity: When possible, label data directly (instead of using legends)

• Why: Help the person interpret the plot (reduce cognitive load, make it enjoyable)
• How: Use pre-attentive attributes (elements that “pop” without searching for them)

Source: Storytelling with data

• Two uses of color:
  • To encode data
  • To guide attention
    • Color is the most useful pre-attentive attribute
    • It also allows for consistency across figures

• Color affect emotion and this is culture-dependent. Some responses are nearly universal
  • Warm colors –> alive/alert
  • Blue colors –> calming/focus

• More information: https://blog.datawrapper.de/which-color-scale-to-use-in-data-vis/ https://davidmathlogic.com/colorblind/#%23D81B60-%231E88E5-%23FFC107-%23004D40

• In addition of highlighting, colours can be used to:
  • Represent categories (no more than 4-5 colors)
  • Represent quantitative values:
    • Only if necessary (i.e. you need to use the x and y axis for more important variables)
    • Not accurate (still okay if you only want to show trends)

• The same shade of gray will be perceived very differently depending on whether it is against a darker background or a lighter one.

• The same shade of gray will be perceived very differently depending on whether it is against a darker background or a lighter one.
• In addition, we are better at distinguishing darker shades than we are at distinguishing lighter ones.

• Colors are a mix of
  • luminance or lightness: (relative) brightness,
  • hue: amount of red, green, blue, and
  • chroma or saturation: intensity or vividness of the color
• To represent quantitative values: Want mappings from data to color that are perceptually uniform
• Default palettes available in R (e.g. ggplot2 or in the package Rcolorbrewer) are perceptually uniform.

Source: https://nightingaledvs.com/color-in-a-perceptual-uniform-way/

• Think about the main message of your visualization
• Think about the way the reader will interact with the plot
  • Use color and other pre-attentive attributes to draw attention to the important parts
  • We read plots in a Z-shaped flow: top-left to top-right to bottom-left to bottom-right

• Type of plot?
• Design principles?

• Data visualization
  • Efficient and effective to show data —> Reduce cognitive load
  • Sticking to basic principles helps (e.g., by using ggplot).

• Why do we want to reduce cognitive load
  • More willing to read your report
  • More likely to understand the data/results
  • More willing to accept the results
  • More likely to remember them (and you)

• When constructing a graphic, consider the following:
  • What is the main message and the indented audience?
  • What are: aesthetics, geom, scale, facets, transformation, coordinate system?
  • Are your most important variables mapped to unbiased channels (length/position)?
  • Can you remove clutter (increase data/ink)?
  • Are you guiding the reader via preattentive attributes to show the main message?
  • Do I visualize the data in a honest way (e.g., am I including the context in a time series)?

How to judge if your model is good: model fit and cross-validation.