The last scenario I want to talk about is a situation where you need to visualize from a given dataset, different aspects of this dataset or if you wish different sets of attributes of these datasets and you can't really fit all of them together in one single visual representation. While this is a very general problem, I often find that this problem arises when the facets that you want to visualize, pertain either to, what something is, who Something is, where something is and when something is. So more precisely, if you have temporal information and then you have spatial information and then you have information about the actual objects that you are visualizing with multiple attributes, you typically have different facets of a dataset that you want to visualize at the same time. Often the problem is that it's really hard to integrate all of them, in one single visual representation. So, how do you solve this problem? Again, the solution is to use multiple views. In this case, what you can do, is to place every single facet in a different view and then link them together using interaction. So one possible solution, for instance, if you have all these types of information, say what something is, when something happens and where it is located, you can in principle create three separate views and then link them together. Let me give you a specific example of how these may look like. Here I created another small interface or dashboard that integrates exactly these three types of information. In this dataset, we have information about flights that are leaving from the three main airports around New York City and we have information about when they leave, we have information about what the destination is, and we have information about what is the main carrier of this flight. This is exactly what you see in this interface. So here, in this view, you see information about where, so what is the destination? This view here, these timeline, depicted with a bar chart, is about the time of the day when they leave and here is about what. So in particular in this case, what is the carrier of the flights that are leaving and going to these destinations? Now what is interesting is that since these views are linked, you can use them to filter out information by selecting some of their elements in one view and see how this selection propagates in the other views. Even before that, you can already notice that even without using any form of interaction, you can already extract very useful information about these datasets. So we can see where the flights are distributed in terms of destination. Overall, we can see a little bit of our time pattern in the temporal view. So we can see that there are quite large number of flights between six and eight, then it goes down and then it slowly goes up again up to 5.00 p.m. and after 5.00 p.m., it goes down and after 9.00 p.m., there are very few flights and there's basically nothing after 11.00 p.m. We can also see that, most of the flights are for these carriers and there are very few flights for these carriers at the bottom. So, again, even just without interaction and linking by presenting these three views together, we can extract useful information about the same data sets. We have three different facets. But now what is really powerful, is that we can interact with any of these views, to see what happens in the others. So for instance, let's say I want to see where the main carrier that we have here, which is United Airlines, where the main destinations are and whether there is a difference in time patterns. Well I just select that, and now I can see that United Airlines is serving only a subset of the destinations that we had in the original image. In particular, there are many flights in Florida, many flights in California, many flights in Chicago, many flights in, I'm not sure, I think this is Houston and something else here. Okay. Let's see what happens if I select the next carriers. So, this is JetBlue. So, when I select this one, I can see the pattern changes considerably. So let me go back to United Airlines. This is what is served by United Airlines and this is what is served by JetBlue. It's considerably different, there are way fewer flights in California and the rest of the US, but there are quite a lot in Florida and around New York City. You can see that also the time hour patterns change when I go from the overall pattern to the United Airlines pattern and to the JetBlue pattern and so on. Of course I could go on forever and select other carriers or select specific regions and so on. Note that, so far I have been filtering the data by selecting specific carriers, but I can in principle also select objects in the map and see how these are reflected in the other views. Let me do that. So, what are the carriers that are serving Florida? So, I'll just select the data points here and as you can see, I have a subset of the carriers, that are popping up and I also have the time of the day when flights towards Florida are available. I can do the same for California. So again, there is only a subset of the carriers and there is a different shape for the time of the day when these flights are leaving. So, I hope I demonstrated that this is really powerful and useful. Keep in mind these three main patterns, as I said at the beginning they are not necessarily exhaustive, but they do cover three very common situations. So every time you find yourself into the situations, you know that they may be solved by using multiple linked views with interaction.
