A2-SethHorrigan
From CS294-10 Visualization Fa08
Contents |
Data Source
To assist me in a class project in the Spring semester, Nick Yee provided me access to part of his raw data from the Daedalus project. Specifically, he provided me with anonymous questionnaire answers from 3251 different respondents. These answers ranged from age to how often their work life has suffered from their game play. In the spring, I coded, cleaned, and formatted all of this data to work with Stata. As such, I could easily export the data from Stata in whatever format I wish. I exported the data as labeled numeric encodings presented as a standard CSV file. With this, I could easily import the data into Tableau. Unfortunately, some of the data did not start at zero, and none of the actual labels were exported correctly, so I was forced to relabel and zero some things.
I chose to work with Tableau to begin because I had already seen a little of the interface in the in-class presentation. In this particular data set, it is not clear what variables should be measures and which should be dimensions. In fact, they can vary, and as they do, new visualizations emerge.
Initial Question
How does age of the players affect their responses to specific questions
- How does age relate to play time per week
- How does age relate to the importance the respondents attach to using the game as escape
- etc...
Process
Initially, I put age on one axis and hours on the other. It looked like a trend immediately emerged, then I noticed that it was comparing the sum of hours.
When I changed this to average, the bulge disappeared. It seems that there are just more players around specific ages.
I tried encoding the number of players at a specific age with the age axis. My first few attempts failed to produce anything useful.
Then I got it right, but I did not get anything really useful there, except to show that there is an abnormally large number of respondents who say they are 18 years of age, and otherwise it the number of players rapidly increases from 17 to age 21, where it begins to steadily decline to age 40. It looks fairly similar to a normal curve actually.
I tried to figure out what else might be interesting. I wanted to include gender with age, and had a bunch of false starts. These failures were mainly due to me learning how to use Tableau more than anything else.
I decided to look at a different variable - whether or no the respondents had ever flirted in-game with someone they did not know in "real life". It was very clear that females were much more likely to have done so (the actual data was encoded as 2=no, 3=yes, so I created a derived variable called "FlirtNormalized" that subtracted 2 from each answer to allow averaging from a zero-point).
With the image above, I got closer to something interesting. With the genders laid out side-by-side, and then dividing responses by age, you can see whether age has a noticeable effect on the flirtation in-game. Unfortunately, since the likelihood is averaged from values ranging between 0 and 1, you are left with spikes and gaps where there is only one or very few respondents at a specific age.
With this iteration, I combined ages into a derived variable that provided 6-year age ranges (although 48 and over is treated as one group since the data becomes especially sparse up there). The result is a more readable visual depiction of the effect of age on flirtation in-game. It is readily apparent that there is a difference in the behavior of the two genders (although the actual cause of the difference is conjecture). There is no clear trend in the effect of age. So I lastly decided to start adding other variables to see if there were effects there.
It appeared that certain variables were affected by age and gender, while others were not. The graphic was supposed to employ Tufte's small multiples to allow comparisons and conclusions among diverse but similar variables and the possible relevant independent variables. Unfortunately, big blocky bars make it difficult to notice trends and compare among the various graphs. Especially with the vertical layout such as it is, there is almost a vibration effect when trying to focus on changes in just one of the sections.
The rest of my work was focused on cleaning up the graph and providing an easy-to-read comparison between the various sections. Although a line-chart is slightly misleading (by implying continuity in a discrete data set), it is readable and accurately conveys the information in a manner that other charts offered by tableau could not. A single, stacked bar chart might have been able to more accurately and concisely convey the information, but it would not allow the "small multiples" comparisons that make this image interesting.
I also played with color, but eventually settled on a blue that was quite close to the default, as it drew the least attention to the color itself while still differentiating the graph from the grid.
Final Product
A visual comparison of the effect of age and gender on aspects of gameplay in MMOGs
As the reader can see, age and gender can have unexpected effects on how individuals engage the game worlds in massively multiplayer online games. At the top, we see a comparison of age and gender with reported importance of actual game-play mechanics, such as getting to know others or making a unique character. Interestingly, there are no distinct differences due to age or gender. On this scale of 0 to 5, in both cases, both aspects of game play were, on average, fairly important to players of all ages and genders. Similarly, the number of hours played per week and the likelihood that the respondent has ever played the game for 10 or more straight hours show variations respecting age and gender but nothing hugely dramatic. There are, however, factors where gender plays a distinctive role. When asked whether they have ever flirted in-game with someone they did not know in "real life", female respondents were much more likely answer affirmatively. Interestingly, age did not play a definitive role in flirting in-game. There is a slight dip in the female population's flirting in the late twenties, but there is no major drop-off as age increases in general. Also, when asked about the importance of defeating other characters in the game (by killing or dominating them), the positive response is fairly high among younger males, but drops off dramatically as they age, and it remains very low for females of all ages.
