Now, I want to use the principles that I have introduced in the previous videos to give you an example of how they can be used for evaluating these visualizations. And I'm going to start with one example. This graph, if you see here, has been published a few years back in the Wired Magazine. And the main purpose of this graph is to show what are the main courses of death in the world. And more precisely, let me try to describe what you can see from this graph and now it's been built. Every single rectangle that you see here represents one specific cause. And the area of the rectangle represents a quantity that is called years of life lost. Which is basically a measure that is used instead of actually showing how many people die for a certain cause, it's more about how many years of life lost are caused by this disease. Why do they do that? Because they want to give more weight to cases when people die young or younger. So size represents years of life lost, color represents three different categories. So injuries and other two categories. And finally, as you can see, within the three color use, yellow, purple, and green, you have different color intensities. And color intensity, as you can see from the legend on the bottom represents the amount of change from one year to another. Okay, so let me summarize. Size of the rectangles represents years of life lost for that specific cause, color intensity represents the change from one year to another and the color yellow represents groups of causes. So what are the issues with this graph? Can we actually improve the representation of this graph? Well, yes, I think we can improve it in several ways. So one thing that stands out, first of all, is that this graph has been created with a 3D projection. And of course since there's a 3D projection, every single shape and the rectangles are actually somewhat skewed by the perspective. And this makes the comparison of these values really, really hard to make. In addition to that, even if we didn't have this 3D visual representation, as we know by what we have seen in the previous modules, area size is not a particularly good channel to represent quantity. And even harder here, you have rectangles that have different proportions between the width and height, which makes it even harder to compare their areas. So that's another problem. Finally, another problem that I see here is that the color encoding that is used to represent change is not particularly effective. Why? Because it's hard to actually distinguish positive changes from negative changes. So when you look at the color intensities, it's very hard to figure out which rectangles represent positive changes and which rectangles represent negative changes. Which is actually a very, very important piece of information because when you look at the whole set of causes, you want to probably quickly detect where the positive changes and where the negative changes are, where improvements basically have been. So that's very important and it's very hard to decode from this graph. Finally, another little problem here is that since some rectangles are very small, it's very hard to show labels that tell you, communicate to you, what this disease is. Okay, so there are a number of issues here. Now, I'm going to show you one possible solution, and I've taken this exercise and this solution from the blog called perception edge, which is actually the website and the blog written by Stephen Few, who's a very popular visualization expert and designer. And this is his redesign of these charts. How did he redesign the chart? So he decided to create three aligned bar charts where one represents the years of life lost for every single cause. The second one represents the amount of change, and the third one represents the absolute values of deaths caused by that cause. Every bar, also as a caller that represents the three categories that we have in the previous graph. What are the advantages of this solution? Well, it basically solves all the issues that I've described in the previous chart, in the previous version. Let me go through them. So comparing the causes, the amount of the ears of life lost across causes now it's very easy, because comparing the height or length of the bar is much, much more effective. Second, since we are using bar charts, it's way, way easier and it's possible to show the labels for all the diseases, which is very useful. Third, by focusing on the middle chart, we can very, very quickly grasp which disease causes have improved or have gotten worse. Now it's very easy. If you see a bar on the right or a bar on the left, you can very easily distinguish the two cases. Not only that, now you can basically correlate the color of the bars to the direction in which they go. So for now, for instance, you can see that all the grey ones go in one direction, right? And the blue ones, some of them go on the right and some of them go on the left and so on. So all these visual differences are so much easier to make. So this is an example of how using some of the notions that I yet introduced in the previous modules can be used to criticize an existing graph and also redesign it in a way that we can make it more effective.
