Review: Visual Display of Quantitative Information

12/29/14

I bought this book about three years ago, I believe. A good friend of mine told me that one of my favorite professors considered this book to contain incredibly important concepts. After reading the book, I think I agree. I finished this book about three weeks ago. This book, like the one I read before it, started with a plan and ended with me just enjoying the read.

The Original Plan

I want to get good at graphing data using d3.js, and this is a book about presenting data clearly and accurately to an audience you respect. So, I thought it would be useful to ingrain my reading of Visual Display by implementing its principles with d3.js exercises.

I didn't do this. At all. I have still done very little with the d3.js library. I will make time for it eventually.

Enjoyable Reading

I was hooked on this book from the beginning, basically. It begins with a history of charts, which really is just fascinating. Like, I mean, I understand that it may not seem interesting to most people, but just think about it: how much do you know about the history of how anything came about? How much do you even think about it?

I had never really thought about the history of charts until reading this book. It's suprisingly short (charts are a new thing), and a few simple mutations from a basic map. In fact, maps are literally what the first charts were. Maps are much older than any other type of chart.

I don't remember the exact ordering of types of charts that grew from the idea of maps, but time-series were among the first types of abstract charts. It made seeing trends against time much easy to see than bare numbers.

Really, the whole book is devoted to how raw numbers could be converted to more digestible forms to strengthen our ability to understand what the numbers really mean. Sometimes, the book argues that the raw numbers themselves are the clearest means of showing the data, but othertimes even presenting raw numbers could be misleading and lead to incorrect conclusions.

The purpose of data graphics is to have the data tell as complete a story as can be told from the data, and let the people interpret that the best they can. Do not spice up the graphics, do not dumb it down, do not tell them what to think. Do not assume your audience is stupid. Respect them and give them the information necessary to form their own informed opinions.

The second section of the book is full of examples precisely showing what lying with graphics looks like. It points out what the data actually are, and why the graphics are painting an inaccurate representation of that data. The author spoke of two kinds of lying:

• Lies of direction
• Lies in magnitude

Most examples in the book were lies of magnitude, but a few were lies in direction. I found this typing of lies to be incredibly useful. Lies in direction are obvious, they are simply the opposite of true. They are much easier to recognize and point out how they are wrong. Lies in magnitude, however, are much more nefarious. They aren't the opposite of true, but they still aren't correct. They often hyperbolize or sensationalize the story, and people are often quick to defend exaggerations. "They aren't completely dishonest," is the usual argument.

The rest of the book was about making use out of every bit of ink used in the generation of graphics to better represent the data. It was here that I should've tried to make d3 graphs that implement the book's ideas, but, you know, I can always do that later. No big deal.

Recommended Reading

Now, at this point, I would normally wholly recommend reading this book, even if you don't like or aren't really interested in charts. I wasn't that interested when I started, but I still enjoyed reading this book and grew a greater appreciation of charts during my reading. It also points out common chart manipulations of data, which could very well protect you from others attempting to lie with data.

But, if you are actually interested in charts, and making better charts that tell a clear and accurate story, I'm going to recommend a book I haven't read yet (but will soon).

After reading this book, I looked at its one-star reviews, and found this one to potentially be a good criticism of the book, and it recommended instead The Elements of Graphing Data.

That book's first review starts with:

So, without even reading the other book, I'll say this:

Want a good read on charts? Their history, their use, how lies are made with them, and how they could be more efficient and clear? Read The Visual Display of Quantitative Information.

Want to learn how to make good graphics? Read The Elements of Graphing Data. That's what I plan on doing sometime next year.