CS-5630/6630 | Visualization | Fall 2014
Week 10: A Banner Year for Policy Spending - Printable Version

+- CS-5630/6630 | Visualization | Fall 2014 (http://www.eng.utah.edu/~abigelow/bulletinBoard)
+-- Forum: CS-5630/6630 | Visualization | Fall 2014 (/forumdisplay.php?fid=1)
+--- Forum: Design Critiques (/forumdisplay.php?fid=3)
+--- Thread: Week 10: A Banner Year for Policy Spending (/showthread.php?tid=70)

Week 10: A Banner Year for Policy Spending - Chaomeng - 11-01-2014 08:16 PM

Banner Year for Policy Spending


This visualization represents the distribution of total amount of money spending on the political election by Democrats and Republicans. It uses a geographical map to represent all region in the United State, and use a circle on the map to represent the amount of money spending in the election on that region. It uses two channel to encode this visualization, the size and the color, where size indicates the amount of money spending for both parties, and the color indicates which party spends more money on that region. The higher percentage of money spent by one party, the deeper of that party’s color would be. One interesting feature here is that it uses the two very reprehensive colors, red and blue, for two parties, which makes it easy to recognize on the map.

On the top of the visualization, the total money are divided into three categories, by candidates committees, by party and by independent groups. This interaction allows you to change visualization and check the detail distribution of money spent from different categories, which is a useful interaction. On the right, the author use a legend to give the total amount of money spend on these categories, and some explanation about some information for this visualization, like the top investment by two parties.

Although I think this is a functional and readable visualization, there is still some issue for this visualization I think could improve. First, there is a problem by using circle to represent money spent on different region. Although it could use size to express its total amount, but circle overlapping with each other makes it hard to read. Especially in the Northeastern part of the map, too much circle shows and overlaps with each other on that small regions, which make it almost impossible to gather where these money belongs to. Another issue is that it is hard to trace which region one circle come from. From the explanation we know policy money are calculated by different regions in state. But the circle makes it hard to trace where that circle belong to which state, especially when some circles are big and cover multiple regions. So it would be better to show a tab and give some information on each circle when mouse click on that circle.

RE: Week 10: A Banner Year for Policy Spending - lzhang - 11-02-2014 04:28 PM

The critique given by Chaomeng is through and accurate.

As mentioned by Chaomeng, the dataset of this visualization is of type map.

The data is the political spending in 109 districts, thus is discrete. The data is treated as points and represented by circles. Since the spending is quantitative, this visualization use size to encode the spending information: the more spending is, the larger the circle is.

This visualization also differentiates between Democratic and Republican. Since this attribute is categorical, the visualization uses hue to represent them. Moreover, saturation is used to represent the percentage of spending by the dominating party: the higher the percentage is, the more saturated the color is. Consequently, we see a diverging colormap in the lower-left corner.

The visualization also uses multiview to convey detailed information of the spending. There are 4 views in total. The first one is for total spending, and the rest three are three sources of money being raised---from candidates committees, from party, from independent groups. These three views demonstrate subsets of the original data set.

I agree with Chaomeng that the size encoding for spending cause problems in this visualization, especially in the region where the area of the states is relatively small like the northeastern part. The intersections of the circles are misleading. According to the description of this visualization, these intersections are purely caused by design and do not mean anything. However, the intersection parts even inherit colors: blue and blue make it blue with higher saturation, and blue and red make it dark gray. These colors may or may not be in the color map. But they do cause confusions about their meaning (they actually mean nothing). As Chaomeng suggested, we may solve this problem by using interactions: for example, the total spending is popped out when mouse is over the region.

RE: Week 10: A Banner Year for Policy Spending - hehe - 11-02-2014 04:45 PM

I agree with Chao's opinion for the most part. The legend describes very clear and different categories do its job well. However, I don't really like color selected for republican and democrat. Because when overlapping occurs, it is hard to identify which is which. For example, in the Total Spending category, on the right corner of the map, there is a black thing and so difficult for me to tell where is it come from. Another thing that I think it could be improved is to apply interactivity on the map. Somehow, I really like interactivity. I believe an interactive visualization would provide more information to users and would be more expressive and effective. I am also curious that why only 109 districts discussed in this visualization. Can these districts present the majority opinion?

RE: Week 10: A Banner Year for Policy Spending - Matthew Turner - 11-02-2014 04:56 PM

I definitely agree with some of the problems discussed in the previous posts. The choice of a diverging color scale to allow viewers to identify parties was good, but the way that the colors mix when circles overlap could lead to misleading interpretations of the underlying data. Furthermore, we have discussed that size and color are not separable encoding channels, so viewers might perceive a region with a very strong bias for one particular party's spending as larger than an equally sized region with a fainter color due to competitive spending from both parties.

I like the ability to switch between the different sources of spending, but I wonder if some alternative techniques could yield interesting results. For instance, using a small multiples layout could lead to easier comparison. Another example could be making a more elaborate mark for total spending, e.g. the circle could be split into a pie chart with the various sources.

A design decision that I found curious was the choice to integrate total spending for both parties into one symbol and then shade that symbol based on which party spent more. If the intended task for viewers is to compare spending by party (as it probably would be), why not encode the data for each party separately?

Overall, I think this visualization's encoding scheme is pretty flawed. However, there are some interesting results that are visible and the annotations on the side help direct users to regions of interest, so it is still possible to come up with meaningful observations.

RE: Week 10: A Banner Year for Policy Spending - u0909249 - 11-02-2014 06:50 PM

For starters, shishir succinctly summed up the "At the National Conventions, the Words They Used" visualization. As some others that have responded have mentioned as well as shishir, as neat a feature to move the bubbles around, there really seems not point besides making the visualization more interactive & fun to use. It is also interesting the semantic approach to the clicking a bubble then having more specific information display at the bottom for the words selected as it pertains to each party. Not only that, but keeping the hues based on party for highlighting the word is a nice affect as well. Overall, the visualization is enjoyable & interactive just as shishir states.

RE: Week 10: A Banner Year for Policy Spending - aseem2691 - 11-02-2014 07:43 PM

Chaomeng has done a good job of explaining the visualization.

I liked the interactivity used in the visualization which makes more sense than to represent all the data in a single map. The map key and the explanation of the information helped in understanding the finer details of the spending.

For me the main problem as discussed in the previous post was the colors used and the sizes of the circle used to describe the amount and the party which spent the money. Firstly the colors colors could have more brighter and complementary with shade more distinguishable. The circles should not have been overlapping. The designer could have used different shapes for different scales of spending. I also with Matt Turner that the designer could have different encoding for different parties which could have simplified the visualization.

RE: Week 10: A Banner Year for Policy Spending - wgormley - 11-02-2014 09:47 PM

First, I agree that the visualization does some things right. Limiting the number of color bins makes it easy to distinguish the different numerical amounts. The interactivity adds a lot, allowing derived data to be seen clearly. But I also think this visualization fails in a few ways. Everything west of the Mississippi is easy to read, but east of there, especially in the north east is almost impossible to read. This makes data abstraction difficult, and severely cripples the overall effectiveness. Also, any overlapping seems to produce inaccurate data. The transparency means new colors are created, that do not seem to fit with in a color bin. Good idea, just bad execution.

RE: Week 10: A Banner Year for Policy Spending - t.li - 11-02-2014 10:31 PM

I agree with Chaomeng's point on the ineffectivity of the size channel and color channel.

- I think the reason why the author picked size as a channel is because states in the northeastern area are relatively smaller than midwest. So, to make marks comparable, there must be a common baseline, which is, intuitively, the size. This choice has its own justification since solely using saturation level for each state is also indistinguishable for densely distributed small regions. However, overlapped colors make things hard to read (as said by above posts)

- I would recommend to use distortion. As long as the distortion doesn't make the US map unrecognizable, states can be shown with relatively equivalent areas. Given this baseline, saturation can be used to indicate the amount of money spent.

RE: Week 10: A Banner Year for Policy Spending - Prashant - 11-03-2014 12:01 AM

Chao meng provided an excellent summary critique for the above visualization. I would like to add couple of additional features which could have assisted in improving the overall visualization design.

1. Display small areas effectively : In order to display the expenditure in smaller regions effectively, A nice way would be to be able to select an individual state or perhaps add a functionality to select multiple states . Based on the selection we could implement zooming into the highlighted section of the map. This would greatly assist to track the progress in smaller regions as the map is mainly targeted towards summarized version of data.

2. Provision to view individual party expenses : I agree with the visual design of the map, the designer has neatly classified data into smaller data groups and plotted them on the map . However in case if one wishes to view independent data expenditure of either parties , having a small button which could help viewers toggle between both either party expenses would prove quite beneficial.

RE: Week 10: A Banner Year for Policy Spending - MukteshKhole - 11-03-2014 06:17 PM

I liked the use of red and blue colors to highlight the spending of the two parties which makes it easier to distinguish. Another interesting aspect is that the visualization allows to select the category in which the money was spent. The overlapping circles do make it problem to correctly identify the information being conveyed, this could have been improved by allowing user to filter the region and the party as well. Or create an overlay for smaller regions and display the overlapping information on it.