Thread Closed 
Week 3: James Bond over the Years
09-09-2014, 07:27 PM
Post: #1
Week 3: James Bond over the Years
[Image: 20121020_bkc276_0.png]

Source: http://www.economist.com/news/books-and-...-you-think

This visualization depicts a data set portraying the average number of occurrences of routine events that James Bond performs during a film for a given actor. This average is based on the number of occurrences per each James Bond film that a given actor starred in.

A stacked bar chart has been chosen to represent this data. Following suit with Munzner, I will denote the mark type for this chart as a line as I feel no information is encoded in the width of the bar.

On the Y-axis, we have a set of several quantitative attributes encoded using both the color (hue) and vertical spatial position channels. These attributes include “Martinis” (the number of martinis consumed), “Conquests” (the number of women seduced), and “Kills” (the number of enemies killed). The intended semantic of these attributes is that they comprise the routine events that makeup a James Bond film. In regards to their types, each of the 3 attributes are quantities representing the average number of times each attribute occurred during each James Bond film that a particular actor starred in.

On the X-axis, we have several categorical attributes encoded using the horizontal spatial position channel. These attributes include the names and picture of each actor who played James Bond during a specific time period (which is not used for attribute ordering).

Overall, I find this to be quite an interesting visualization that effectively portrays the varying amounts of each key event found in a James Bond film. It is interesting to discover both the spike in alcohol consumption during Daniel Craig's portrayal and killing during Pierce Brosnan's portrayal. I believe that these are two of the main points that the visualization attempts to draw focus to (and effectively does so).

In regards to aspects that worked well, I very much like the hues chosen to represent each quantitative attribute as they are easy on the eyes and different enough from one another to maintain separability. I also think that the icons next to the quantitative attributes and actor portraits work well from the standpoint that they allow one to ignore axis labels if they are familiar with each image.

In regards to aspects that didn't work well, I disagree with the decision to use a stacked bar chart. While it allows for quick judgments to be made regarding the cumulative occurrence of all 3 quantitative attributes, it makes it difficult to make judgments based on only the "Martinis" or "Conquests" attributes. This holds due to Weber's Law.

Rather, an improvement that I would propose is using a grouped bar chart instead where each quantitative attribute shares a common baseline on the X-axis. This would increase the clarity of the visualization and allow for comparisons to be made quickly between individual quantitative attributes.
Find all posts by this user
09-10-2014, 08:47 AM (This post was last modified: 09-10-2014 08:48 AM by Khalid.)
Post: #2
RE: Week 3: James Bond over the Years
I agree with u0923385 that stack bars are not the best choice to represent this data. A better choice might have been a line chart graph, which will clearly show the differences and trends between the different James bond actors.

A simple color improvement that can be made is changing the bar chart background color to plain white instead of a very light blue color and consequently the chart horizontal lines to black color instead of white. This change will make the legend color saturations more distinguishable and the stacked bar chart edges easily readable.
Find all posts by this user
09-10-2014, 08:07 PM
Post: #3
RE: Week 3: James Bond over the Years
While this chart is ordered by combined frequency of martinis, conquests, and kills, I think it would be interesting to order by time. At first glance I thought, "Oh, violence has gone down over time." But then comparing Brosnan to Dalton I thought, "Oh wait, it's gone up over time." To then realize that the actors aren't ordered by time at all.
Find all posts by this user
09-11-2014, 11:13 AM
Post: #4
RE: Week 3: James Bond over the Years
I agree with aedunn. I expected the x-axis to be ordered by time, because I know that the actors changed over time. Not ordering by time makes it very hard to determine a trend, which would seem to be the purpose of a visualization titled "James Bond over the years".
Find all posts by this user
09-11-2014, 03:39 PM
Post: #5
RE: Week 3: James Bond over the Years
The visualization as the title on graphic reads tries to depict a picture of "James Bond over the years" by depicting three common events/activities associated with the lead role and how it has changed/evolved over different periods of time when different actors played that part. Personally I feel there is a lot that could have been improved while coming up with this visualization :

1. The stacked bars, like all the others have mentioned here is probably a bad choice for this visualization and the reasons have been precisely put by u0923385. I would personally have a side-by-side bar like the one in Tableau to represent this data. It would give a much clear picture about the three different activities over the years and will provide a basis for comparisons and inferring results.

2. The graphic displays the three activities of kills,conquests and martinis on the X-axis. However for a person who does not watch Bond films or someone who hasn't heard about these films, interpreting these activities might not be straightforward. Case in point, there is the conquests with a broken heart symbol. What does it mean ? It could be interpreted in different ways by different people. The visualization could be much better if it had some sort of semantic or contextual information about these three events.

3. I would have to agree with Mr. Edward Tufte here and say that horizontal white lines on the graphic does not do any good and can be removed.

4. I would swap the scale on the x-axis which appears on the right and have it displayed on the left and move the activities/symbol legends to the right. I feel personally this would make it much easier to interpret or find the intermediate values as we are used to having scales on the y-axis towards the left side of the origin and intuitively we can do the computation as opposed to this graphic where it is on the right.

5. I agree that an ordering by time would make the visualization much more useful. I also personally feel if the graphic had displayed the count of Bond movies each actor had played right next or below their timeline, it would give an exact picture of how the activities have evolved and wouldn't lead us to make biased conclusions of how James Bond has changed over the years. This is especially due to the fact that the graphic tries to display the averages and not the exact count.
Find all posts by this user
09-11-2014, 04:25 PM
Post: #6
RE: Week 3: James Bond over the Years
I completely agree with u0923385 on the critique of the graph.
I had problems with the “stacked” bar chart as well. Maybe a side-by-side bar chart would have been clearer. With the color, using blue and different hues is pleasing to the eye. But I am a little skeptical about how it would turn out if it were a line chart as suggested.

In terms of the ordering along the x axis, I agree with aedunn too. I think it would be much more interesting if was based on time. But, adding pictures of the various actors who played Bonds was a good idea I guess.

I also agree with comment on Weber’s law. All in all, I think this is a good critique.
Find all posts by this user
09-11-2014, 08:10 PM
Post: #7
RE: Week 3: James Bond over the Years
Regarding ordering the X-axis by time, I agree that it would be interesting to understand how the role has evolved over time and determine whether there are any interesting trends.

However, one issue that I do see with ordering by time period is that George Lazenby's portrayal falls within Sean Connery's portrayal.

I see two possible ways to approach this depending on the overall message you wish to convey with the visualization. If you wish to maintain true to sequential time, Sean Connery's portrayals could be split into two separate attributes with George Lazenby's portrayal represented between those. If you wish to maintain the "per actor" aspect of this visualization, attributes could be ordered based on start date with Sean Connery's portrayal being representing by only one attribute.

Or, perhaps ordering by time was explored when creating this visualization and the creator didn't discover an interesting trend. This could be why time was neglected and emphasize was placed on ordering by the average number of routine events.
Find all posts by this user
09-11-2014, 08:55 PM
Post: #8
RE: Week 3: James Bond over the Years
This is indeed a great bar chart. I agree with all the points that 'u0923385' made, esp. the one when it was stated that a group bar chart would have been much better.
The average number of kills can be counted here as the base line is in conjunction with the x-axis but it takes a little while to figure out the average number of martinis and the conquests.
However, "I feel" that the role these actors played should have been placed in orderly fashion depending upon the years in which they acted staring from 1969 i.e. having a bar associated with 'George Lazenby' first followed by Sean Connery and ended with Daniel Craig instead of portraying the occurrences of the events in descending order.
Over all, I think this is a good critique.
Find all posts by this user
09-12-2014, 10:07 AM
Post: #9
RE: Week 3: James Bond over the Years
This was a very good critique.

I agree that if you are going to title your visualization "James Bond over the years" then the dates along the x-axis should be chronological so we can see the progression over time.

I also agree that if we are going to maximize effectiveness we need to use points on a common scale for this visualization. That would mean using the grouped bar chart or something similar as suggested in the critique. However, if the author is trying to convey that the Pierce Brosnan era had the greatest total of all three attributes, you could just label the marks with their actual data value so you can do a quick comparison without having to interpolate off of the y-axis.
Find all posts by this user
09-13-2014, 01:37 PM
Post: #10
RE: Week 3: James Bond over the Years
Quote:Overall, I find this to be quite an interesting visualization that effectively portrays the varying amounts of each key event found in a James Bond film.
I only partially agree with this statement, as I personally believe the way the data is portrayed is confusing. At first glance the vertical size of the color encoded data channels, seems to be indicating magnitudes of events rather than actual numbers. An example of this confusion comes directly from u0923385
Quote: It is interesting to discover both the spike in alcohol consumption during Daniel Craig's portrayal and killing during Pierce Brosnan's portrayal.
However, if you look at the actual numbers on the right hand side of the graph you will notice that Pierce Brosnan's alcohol consumption numbers are much higher than Daniel Craig's. Thus, I feel that this graph is intentionally misleading people into believing that alcohol consumption has increased over time. When in fact, it is somewhere in between old bond films and pierce brosnan james bond films.

Quote:...I very much like the hues chosen to represent each quantitative attribute as they are easy on the eyes and different enough from one another to maintain separability.
Agreed, this graph is definitely visually pleasing and easy on the eyes.

Quote:...I disagree with the decision to use a stacked bar chart.
I partially agree with this statement. From a design stand point, it is a huge waste of ink as the data/ink ratio is rather low. However, considering the audience this is targeting, I believe it to be a very effective visualization.

Quote:Rather, an improvement that I would propose is using a grouped bar chart instead...
Agreed, but it seems to me that the data is purposefully trying to be misleading.
Find all posts by this user
Thread Closed 


Forum Jump:


User(s) browsing this thread: 1 Guest(s)