Name: Joel D. Rodriguez-Ortiz
The bottom view is a scatterplot matrix of all the numerical values of the football data set. This view can be panned by dragging with the mouse and zoomed in and out with the mouse scroll wheel. Double-clicking on any of the scatterplots in the matrix will update the (upper) juxtaposed view with the selected scatterplot. Points in this plot can also be clicked to show player details in the corner. The details shown in the corner are filtered by the plot selected from the matrix, they will match the axes shown in the currently selected scatterplot.
The original intent of this idea was implementing juxtaposed, coordinated, side-by-side (top-to-bottom in this case) views. The views share a subset of the data in an overview/detail type fashion.
For the longest time now I've been wanting to use other attributes from the football data set other than Passing Yards vs Passing Attempts. And scatterplot matrices are a perfect example of the small multiples idiom discussed in CH12. So I thought why not just plot all the attributes of the dataset?!
The idea was that you would use the scatterplot matrix to search for trends of interest. But, by providing the ability to pan, I wouldn't have to limit the number of attributes in the matrix. In essence you were also using it as a bird's eye map, to hone in on a combination of attributes that exhibited a strong relation to each other and then shifting your attention to the detailed view (akin to the CH6 Overview First Details Later rule of thumb). Scatterplot matrices are also an example of small multiples, another idiom discussed in CH12.
The implementation worked more or less as I wanted, but ran into a problem mentioned repeatedly in the chapter: screen real estate (this may have been made worse by working with a bl.ocks.org block). I wound up making the detailed view slightly smaller than I originally wanted to and had to stack them vertically because horizontally they looked odd inside the block. I placed the scatterplot matrix below the detailed view as if they were navigation controls. In the end both views fit into the screen and exercise juxtaposed views as originally intended. The navigation flow should go from searching for trends in the scatterplot matrix, to selecting a scatterplot showing attributes of interest, to further inspecting individual values within the scatterplot.