Customer Reviews

Creating More Effective Graphs

5 star:		(7)
4 star:		(3)
3 star:		(1)
2 star:		(1)
1 star:		(3)

 

 

This is no coffee table book for the numerati. It is practical handbook for anyone who wants to use graphics to help people to understand data. "One graph is more effective than another," Robbins says, "if its quantitative information can be decoded more quickly or more easily by most observers."

Yes, Robbins can teach you how to lie with statistics by using three-dimensional bar charts, how to confuse by using stacked bar charts, and how to obscure by using the plain old pie chart. But her focus is demonstrating truth-telling techniques that promote understanding by providing visual clarity. Some of these techniques were new to me, such as the dot plot, the jittered strip plot, the trellis display, and the box and whisker plot.

What's a box and whisper plot, you ask? Unlike this review, Robbins book is filled with examples. Most two-page spreads show a graphic on the left page, and one or two paragraphs of text on the other. With a few exceptions, the graphics were generated by Robbins herself, which gives the book a consistent look that is pleasing to the eye.

These two-page spreads help make the book fairly modular. If, for example, you are not familiar with logarithms, you can easily skip over those sections. Read the first three chapters to understand her approach. Read more until you get bored. Then skip, skim and scan, looking for the juicy bits.

Here's one "Scales: Must Zero Be Included?" she asks. It depends, she answers, and illustrates her point with a description of guru Edward Tufte during his workshop on presenting data. Tufte gets up on a table, opens a book, and unfolds a long narrow piece of paper some seven or eight feet long. At the top was a small line graph. At the bottom was the zero.

Although Creating More Effective Graphs is not scholarly in style, Robbins gives due credit to her teachers and sources. Scholars will find the book useful for their academic writing, and teachers of courses such as quantitative methods, critical thinking, and media studies will mine its ideas for their course notes. But the book is squarely aimed at practitioners: economists, business writers, journalists, or anybody who has to back up an assertion with hard data, and wants to illustrate that data with an effective graph.

This is a great book for those who really want to make their charts/graphs more effective. Robbins' discussion on visual perception helped me understand why pie charts and stacked bar charts are such poor communication tools. Her dot plot examples convinced me to start using them even though Excel doesn't provide a dot plot chart type. I made my own.

My charts are now much better because of this book! While Tufte convinced me that pie charts were bad, Robbins explained why and showed me how to use dot plots to replace pie charts/stacked bar charts. I picked up a number of other important techniques,including cycle plots and trellis displays. She has excellent advice for Excel users, including a link to a dot plot template for Excel.

Her discussion on trellis displays convinced me that there are other charting tools beside Excel that can be used for multivariate charting. I learned about R (free statistics and graphics package) and have started using it so that I can do some of the trellis charting that Robbins explains so well.

Even if you have Tufte's books, you will learn practical aspects of making effective charts with this book. Robbins shows you how to do what Tufte recommends.

This book gets 5-stars for those persons that have trepidation about really modifying how they do graphing. For the fearful person, no other book is as gentle, yet effective, at convincing you of the inadequacy of your simple (most likely Excel) ways.

The real point of Effective Graphs (both this book and the subject itself), though, is not making graphs just a little bit better. There's much more substance than a minute improvement in one's graphs. The real point is that data behaviour elucidation has two rather distinct paradigms: (1) the statistical inference paradigm (tables of descriptive statistics, parameter estimation, test statistics, hypothesis testing) or (2) the William S. Cleveland Visualization Paradigm (well-done simple graphs as well as graphs plotting more complex or highly-derivative quantities).

In the statistical inference paradigm, what one sees, literally, is only big bunches of numerals -- the depiction of those abstract entities we call numbers. In point of fact, even if you were literally blind, the statistical inference paradigm of data behaviour elucidation would work just as well for you as it would for a sighted person. In diametrically opposite contrast is the William S. Cleveland Visualization paradigm in which you will literally SEE data behaviours. This is what this book is about.

"Creating More Effective Graphs" is about SEEing data behaviours. The book is therefore targeted to anybody who wants to show data behaviour, but especially those folks not in the scientific or statistical worlds -- although, people in those worlds will also find, as the title suggests, very effective ideas when taken to heart. For the more advanced issues of data behaviour formulated as rather advanced statistical questions, you should refer to William S. Cleveland's book, "Visualizing Data", which shows how even advanced statistical questions can be addressed in the visualization paradigm.

The conclusion is this: if you want to make a truly substantive improvement in data behaviour depiction and elucidation by your graphing, then "Creating More Effective Graphs" will be an admirable companion in this endeavor. If you need a reference on the "whole enchilada" of what is available in the visualization paradigm on data behaviour elucidation including the more advanced statistical issues, then you will need William S. Cleveland's "Visualizing Data" book.


A FOOTNOTE ON COLOR: Other reviewers' talk of color lacking in the book are completely misguided. The visualization paradigm can do so extremely much in purely black and white. Fanciness and color is not what visualization of data behaviours is about. (And need it be pointed out that there is the great convenience of transmitting and reproducing your effective graphics in black & white.) Well-done graphics using principles of human graphical perception (see William S. Cleveland's "Elements of Graphing Data") and often advanced computations in order to construct advanced plots and plot ensembles is the heart of the visualization paradigm. But on the topic of color, William S. Cleveland does give some good advice in his "Visualizing Data" book. (Just as an example, we all think that the "rainbow" scale for visualizing temperatures of an object or scene is the obvious right choice. Cleveland explains how such a choice is BAD for a variety of reasons. The right choice is a mere two hues with varying degrees of lightness to encode the temperatures.)

In the world of statistical graphics, there are two big names: Edward Tufte and William Cleveland. Tufte writes for graphic designers who want to learn some statistics. Cleveland writes for statisticsians who want to learn some graphics. All their books are excellent, and the author of this book (Naomi Robbins) stresses that.

But there is another vast audience: People who are neither graphic designers nor statisticians, but who need to make effective graphs. This audience includes businesspeople, academics in many fields, and anyone who needs to present information graphically. In addition, many people need to read graphs.

For all these people, this book is close to ideal and is highly recommended. Robbins writes clearly, chooses good examples, and shows both good and bad graphics (and states why the former are better than the latter).

This book has a very important quality, it is very objective and simple to read.

Most of the book, has a structure that makes it very easy to read: on the left page you have one or two graphs and on the right page just one or two paragraghs explaining and comparing what is shown in the graphs.

The organization of the book is very rational, looking at the table of contents you have a very good summary of what you are going to get.

But....the price is too high for a black and white book on graphs. I would say at half the price this book could sell a lot more.

I liked the approach taken by the author to show a good and a bad example of graphs about the same data. It's about graphs, so there is not much text, but it is making to the point remarks about errors in the bad graphs.

Robbin's book has been grossly over-rated. The first major negative is that there is no discussion of the use of color and no use of color in the book. For a book on effective graphs, it is criminal not to use or cover the impact color - there is no excuse. The second major negative is that this book is little more than a set of graphs on one page followed by a brief one or two paragraph discussion. My criticism here is that there is too little discussion about when to use these graphs or common design flaws when they are used. Business users should be warned that most of the graphs covered in this book were created using the S+ programming language and software and many are not applicable to typical business graphing needs. Finally, the book is excessively overpriced for what it provides. Stephen Few's "Information Dashboard Design" and "Show Me the Numbers" books are both half the price and have 10X better content.

This is a very good book. I refer to it frequently. This book is particularly useful if you are a business reader who wants quick and easy to digest tips for improving the display of data. For the technical practitioner, there is a lot here as well.

The presentation throughout the book is very clean and simple. Perfect for a book that is about making information consumable.

This book is a superb presentation on scientific graphics but it is not ideal for the marketing folks (hence the love and hate reviews). If your goal is to make graphics that can be quickly and accurately digested in scientific circles (poster presentations or scholarly journals) then this book is a must. If you are going for attention grabbing good looking color spreads try other books (begin with slide:ology by Duarte).

These days most researchers are still forced to deal with black and white or gray-scale images for publications and this book will help you render great non-color graphics but with the use of web-presentation and color poster sessions there is a need for a 2nd edition to cover the use of color.

I strongly suspect that people who read this will want to get their hands on a real statistical graphics program like R, S-plus or SAS 9.2 (or newer) to be able to take full advantage of the suggestions but they are not required because with this book and some of the references (with websites given in the book) you can generate almost every graphic in Excel easily (box-whisker plots are a big exception).

The author does not do the Microsoft bashing that you see from other authors who love graphics (Tufte especially) but if you read this book you will understand their disdain for Excel and PowerPoint graphics.

The basic ideas in the book are valid and well worth knowing but the presentation is disappointing. There are many reviews already posted that discuss the lack of color used in the book. I, for one, would have preferred that the graphs had been in color since 90% of the graphs that I do are in color. However, I could have gotten past that if one of the points was on how to make B&W graphs visually appealing. But the author didn't do that. The graphs that were obviously originally created in color were merely reproduced in B&W. To make matters worse, the quality of the graphs is poor. They look like photocopies of photocopies of photocopies. Also, while the book is 8 7/8 inches tall, the top and bottom margins are nearly 3 inches each, leaving only approx 3 inches for actual text on most of the pages! This might have been a worthwhile $15 book but, in my opinion, it is not worth near the $50 they are currently asking for it.