Data Visualization Explained (Part 2): An Introduction to Visual Variables

article in my information visualization sequence. See the earlier article: “Data Visualization Explained: What It Is and Why It Matters.”

So, now you’ve realized the foundational thought of what underlies information visualization and why it’s a vital part of the information science ecosystem. (In case you are not accustomed to this, you should definitely take a look at the article linked above.)

As we mentioned within the earlier article, the core thought of knowledge visualization is discovering an efficient approach to symbolize information of varied varieties in a visible method.

The important thing underlying idea which makes this illustration work is named a visible encoding channel. A visible encoding channel is successfully the means by means of which numerical, textual, or another type of information is translated into a visible mark. The easiest way to consider it’s as a visible function similar to all or a part of your information. Efficient information visualizations typically use a number of visible encoding channels for various points of the information.

On this second article, we’ll dive into the main points of visible encoding channels and acquire some apply breaking down a fancy visualization into its element elements. This can put together you for designing your individual visualizations within the close to future.

Introduction to Visible Variables

In his 1967 work, The Semiology of Graphics, French cartographer Jacques Bertin outlined seven “retinal” variables, named as such as a result of the human eye’s retina is delicate to them [1]:

1. Place (such because the coordinates on a graph)
2. Dimension
3. Form
4. Shade hue
5. Shade worth (lightness to darkness)
6. Orientation
7. Texture

Though Bertin printed his work many years in the past, his visible variables stay a wonderful guideline for contemporary information visualization design. Within the early phases of creating a visualization, it’s good apply to overview the visible variables obtainable and decide which of them to make use of for particular variables within the information.

This could be a complicated idea and is extra simply understood with an instance. The graphic beneath, typically thought-about a masterful software of visualization, was designed and drawn by Charles Minard. It depicts Napoleon’s failed invasion of Russia.

This can be a simplified and translated model of the map to ease readability; for the unique, see here [2].

What completely different visible variables are getting used within the graphic above? (Trace: There are fairly a couple of.) As an train, get out a pen and paper and attempt to decide this your self. We’ll stroll by means of it intimately in a bit.

Maximizing Effectiveness of Visible Variables

The very best visible variable to make use of for a selected visualization is dependent upon the information. Right here, we are going to have a look at three several types of information:

1. Quantitative: Numerical information with a pure ordering that’s appropriate for mathematical operations (i.e., it is sensible so as to add/subtract/multiply/divide particular person information values). For instance, wage and age are quantitative variables.
2. Ordinal: Categorical information (i.e., non-numerical information which may tackle a set variety of values) that also has a pure ordering. If in case you have ever taken a survey with reply decisions resembling “Strongly Agree,” “Agree,” “Impartial,” “Disagree,” and “Strongly Disagree,” then you will have seen ordinal information in motion. Whereas mathematical operations on this information don’t make sense, numerous values can nonetheless be ordered from “finest” to “worst,” so to talk.
   • This additionally consists of variables that will have an order with out technically being “ranked,” resembling visitors gentle patterns.
3. Nominal: Categorical information which has no pure ordering. An ideal instance of that is colour. Whereas it’s doable to tell apart between completely different colours, they haven’t any pure sequence. (This additionally explains why colour is a superb visible encoding for nominal variables generally, as we’ll see beneath!)

Vital: Simply because a variable is a quantity doesn’t routinely make it quantitative. For instance, zip codes are numbers, however they haven’t any pure ordering, nor can one carry out mathematical operations on them. Thus, zip code is a nominal variable.

The next desk, a variation of 1 designed by visualization specialists Jock D. Mackinlay and Stuart Card, outlines the effectiveness of various visible variables relying on the kind of information [2]:

A couple of key factors about these rankings:

• Place is the most suitable choice for all variable varieties. For instance, a bar graph with names on the x-axis and blood stress on the y-axis makes use of place for each a nominal variable and quantitative variable, respectively.
• After place, desirability adjustments for every variable kind. That is vital to know as a result of if you’re graphing a number of variables, you’ll finally have to make use of one thing aside from place as a result of it’s already getting used (often on a 2-D graph with two axes).
  • Size is an extension of place, however particularly helpful for quantitative comparisons.
  • Density and saturation are nice for ordinal variables, as your viewers don’t want to find out precise values—they only must see the rankings.
  • Hue and form work nicely for nominal variables, making it straightforward to see categorical variations.
• Some choices are totally crossed out as a result of they merely wouldn’t make sense. For instance, form isn’t a doable encoding selection for quantitative or ordinal variables, as a result of there could be no approach to examine portions or perceive orders.

Now, let’s stroll by means of an instance of the way to break down visible encoding channels intimately.

Minard’s Map: Breaking Down the Variables

Let’s have a look at Minard’s map of Napoleon’s invasion collectively. Right here it’s once more for comfort. This instance is taken from Edward Tufte’s well-known visualization ebook, The Visible Show of Quantitative Data [3].

A cautious examine of this map exhibits Charles Minard’s mastery of visible encoding channels as nothing in need of sensible. His visualization shows six completely different variables:

1. Geographic Location (Quantitative): Place is used to show the situation of Napoleon’s military on a 2-D floor (so that is technically two variables). The invasion started on the left aspect of the map, on the Polish-Russian border. We will additionally see how at occasions, elements of the military department off to completely different places as a part of Napoleon’s technique.
2. Geographic Location (Quantitative): See above.
3. Time (Quantitative): Wanting carefully, we will see that numerous cut-off dates are listed on the chart’s x-axis on the backside of the visualization. Once more, the place is used to show this variable.
4. Temperature (Quantitative): Temperature is plotted in relation to time on the chart beneath the map. Place is used but once more, this time on the y-axis.
5. Variety of Troops Remaining in Military (Quantitative): The width of the form transferring throughout the map represents the variety of troops in Napoleon’s military. It’s clear that because the invasion progressed, Napoleon’s military turned smaller and smaller. They finally returned to Poland with solely 10,000 dwelling troopers out of an preliminary 422,000.
6. Route of the Military’s Motion (Nominal): Shade is used to depict the course wherein the military strikes at numerous positions. The beige/tan colour (white within the simplified picture we’ve above) signifies the military’s motion towards Moscow, and the black colour signifies its retreat again into Poland.

In his ebook [3], Tufte refers to Minard’s map as probably “the perfect statistical graphic ever drawn.” Learning it may well encourage us to plot intelligent methods to encode our personal information visually.

Ultimate Ideas and Wanting Ahead

With this second article, you’ve realized the foundational thought behind visualization design: visible encoding channels. As you mirror on what you’ve realized, hold the next key factors in thoughts:

• The selection of visible encoding channel can typically make or break a visualization. You may need a superbly designed graphic, but when the visible encoding channels are exhausting to interpret, your viewers gained’t know what you’re attempting to say.
• Place reigns supreme for all variable varieties, however there may be restricted area in a 2-D atmosphere. As such, think twice about which variables you show with place; they’ll typically be a very powerful ones.
• Check out completely different designs! There isn’t a “one” good resolution. Quite, you have to revise and reiterate till you attain a passable level.

Within the subsequent article, we’ll discuss vital suggestions for visualization design and the way strategies have advanced and expanded during the last a number of many years. Till then.

References

[1] Semiology of Graphics, Jacques Bertin (translated by J. Ronald Eastman)
[2] https://ageofrevolution.org/200-object/flow-map-of-napoleons-invasion-of-russia/
[2] Readings in information visualization: using vision to think (Card, Mackinlay, and Shneiderman)
[3] The Visual Display of Quantitative Information, Edward Tufte