The 6 Principles of Quality Data Visualizations (Introduction)

90% of the world’s data has been created in the past 2 years. Big data is here; and it is here to stay.

And along with this increase in the world’s data, there has been a commensurate increase in the demand to communicate it. Here is where data visualizations come in.

Data visualizations are simply visual representations of data. These include charts, maps, graphs, etc. The primary reason to visualize data is to make your analyses more intuitive to the audience. A quality visualization can also make or break a presentation. I’ve seen it happen. Clunky graphs with hard-to-read labels can leave the audience confused, whereas a crisp, fluent display can really stand out and become memorable. There is even some evidence that neuroscience findings are considered more believable if they are represented on a colorful image of a brain compared to just shown in a table of statistical results.


In a series of upcoming blog entries, I am going to examine 6 fundamental principles behind great data visualizations. These principles are based on Edward Tufte’s book Beautiful Evidence (which I – along with many others – highly recommend). If you follow these principles, you will be making outstanding visualizations in no time!

In the next entry, I will focus on the first of these 6 principles: namely, that a good visual display should show comparisons, contrasts, and/or differences.

In the entry I will draw a bit from my background in perceptual psychology and cognition. I will first focus on the underlying reasons behind why this technique is effective, and then I will demonstrate how to use this technique to make effective visualizations.

Stay tuned!

Advertisements

My Second Data Visualization

This is my second visualization. I am particularly proud of the “story” I told here with this data. It can get complicated, but I think it is interesting. The idea came to me after reading an opinion piece in the New York Times that detailed the many benefits of increasing access to education for women. Now the focus of this article was on developing countries, but I thought perhaps there was already data on this topic from the U.S.  Turns out there was. So I sought out the data, did a little bit of an exploratory analysis, found some interesting patterns, and turned it into a graph using R. Here is what I found:

visualization_gender_parity.png

At four different levels of education, this chart tracks the ratio of male to female average income over time and the ratio of males to females in the workforce. Thus, on the Y axis, the value of 1.0 represents parity (a 1:1 ratio of male to female).

As you can see, with the exception of those who only obtained a high school diploma, as the ratio of male to female graduates approaches parity, so too does the ratio of male to female income levels. Now, of course, one should be cautious in accepting correlational research for anything more than just correlations. The point of this chart is not to imply an underlying mechanism for the observed effect (e.g. that increasing access to education for women causes more income parity by gender). It is simply showing that these two phenomena are naturally co-occurring.

 

In my next post, I will give some tips on how to make a quality visualization. In short, it involves both a compelling story AND quality aesthetics. I will give some pointers about each of these components, to help you get the most out of your visualizations.

My First Data Visualization

Hi. I have spent the past year and a half “beefing up” my chops in the world of data science. Part of that has included learning some techniques in data visualization.

Here is some work I did “just for fun” using a dataset from the U.S. Census Bureau. The graph shows the patterns of income difference between black and white people in the U.S. over time.

vizualization_race_income

Note: All dollars are scaled to 2013 values to adjust for inflation. Data goes up to 2005, assuming that the 2007 economic crisis presents an historically anomalous condition.

Importantly, it considers income disparity at three different levels of social stratification. For example, the yellow line represents the income disparity between the top 5% of white earners and the top 5% of black earners.

Overall, the chart indicates that in the U.S., historically white earners have made more than black earners. That is perhaps not surprising on its own. However, it is interesting to note that as you from low to high levels of stratification, this disparity increases not only objectively over time (i.e., in terms of the total nominal value), but also relatively over time. This suggests that over time, U.S. income polarization has especially favored white earners, particularly those who are earning the most.

Rocks in Your Head

Today I am going to tell you that you have rocks in your head. But don’t freak out; they are fairly harmless rocks. And in fact, throughout the course of your day, they are useful to you in all sorts of ways. You know when you’re tipping back in your office chair and you can sense that you’re just about to tip a little too far? Well you can thank these rocks in your head for allowing you to re-calibrate and preventing your fall!

They’re called otloliths (Greek for ear stones) and they’re crucial for helping us to maintain a sense of balance. As the etymology would suggest, they’re located in our inner ear (home to our vestibular system) along with a bunch of other stuff that is in there to help us walk, run, score touchdowns, etc. without ever falling over. They respond to both linear acceleration (so moving up or down) and the passive tilting of our heads (for example, when you look up at a skyscraper).

How they work is pretty basic: they’re fixed in a gelatinous material in your ear and surrounded by little tiny hair cells that send messages to your brain – either about your up/downward acceleration or your head tilting. So when you’re in an elevator or looking up at the sky, gravity acts on these rocks and essentially pulls them backwards, exerting pressure on those hair cells and starting a chain of reactions that tells your brain that you’re moving up or that you’re falling back.

Though otoliths are ultimately a key for helping us to stay up on our feet, they can be quite the little tricksters at times, and in some cases, they can even work against our best interest. Case in point: if you’re a pilot.

So imagine you’re a pilot. You’re just taking off from a runway. As the plane first leaves the ground, your otoliths start responding to the linear acceleration. They thus send you a signal of, “hey, you’re moving upwards.” But remember – they also respond to passive tilt. So during this takeoff, as your head naturally tilts backwards, the otolith rocks transduce even more signals from gravity – and what you get is essentially a “double whammy” of sensory information at once, both from leaning backwards and moving upwards. The result is an overrepresented, illusory effect on your perception, particularly about the angle at which the plane is flying (as a result, you end up overestimating the angle).

For pilots, without an awareness of this phenomenon, this illusion can mean life-or-death, and in the past, it often did. Back in the day, many planes were crashing at nighttime because of this “takeoff problem”; pilot trainees, relying on their perceptions alone as a guide for steering the plane, were unable to reconcile what their brain was telling them (i.e., the overestimated angle) from what – in reality – was the actual angle of the plane. The upshot of these early plane crashes, of course, was that during those initial moments of takeoff, pilots had to learn NOT to rely on their intuitive human senses, but rather to put stock in objective measures, such as gauges and -odometers, instead. Luckily for us today, all modern pilots are trained to use such measures.

You may be thinking, why would this happen at all? Our perceptual systems, after all, seem to do a pretty good job of getting us through our days in a relatively safe manner. So why in this takeoff scenario would our perception tell us one thing, but then the odometers tell us another? Well, it’s simple. Through evolution by natural selection, people have come to adopt some incredible capacities, perceptual ones included. Flying a plane, however, is not such a capacity. After all, planes were invented in 1903. That is relatively recent in the ~200,000 some years we have existed as a species. Had there been many analogous scenarios to this “takeoff problem” in the ancestral living space, perhaps we would have developed a mechanism to correct for it. But the fact that this illusion is still built into the perceptual systems of modern-day humans lays testament to the claim that – back in the day – these scenarios must have been relatively uncommon. Natural selection is not perfect. The way that it works is that the organisms which survive are the ones that adapt to the most pressing environmental constraints of the time. The “takeoff problem” was never such a constraint. Thus, developing a corrective mechanism for it would not have conferred enough of an advantage so as to override any of the existing perceptual hardware. So in 2015 we still have the same otoliths that we’ve always had; one’s that are imperfect for sure, but which were good enough at the time (and still pretty much are).

In sum, yes you have rocks in your head, but in most cases, they are there to offer you balance. More importantly, the story of otoliths in pilots is meant to drive home the point that we can not always put faith in our senses. If possible, it is usually wiser to consult facts and to employ some objective measures when we are making decisions.

So there’s my tidbit of sensation and perception for the day. Pretty neat stuff if you ask me!