Buy
Tableau Supports the Next Generation of Engineers by Sponsoring FIRST Robotics
Tableau recently sponsored Wolverine Robotics Team 949 from Bellevue High School in the 2009 FIRST Robotics Competition. It was a good year for Team 949. They were given a Judge's Award this year for succeeding with an ambitious plan to grow the team (they went from a very small team to 32 students and started 2 FIRST Lego teams at the local middle school). They also were one of the top 8 teams in the preliminary seeding rounds at the Seattle Regional competition.
My daughter was co-president of Team 949 this year, which was a priceless educational experience. The students had 6 weeks to design and build their robot with the help of volunteer adult mentors who do science and technology in their day jobs. Content experts are typically very effective teachers. I remember, in particular, the day of the competition when a motor had failed in the robot. My daughter was deep in the robot diagnosing the problem when everyone realized it was time to head to the floor for the next match. She deftly put the robot back together and they made the next match. FIRST (For Inspiration and Recognition of Science and Technology) has the mission to inspire young people to be science and technology leaders, by engaging them in exciting mentor-based programs that build science, engineering and technology skills, that inspire innovation, and that foster well-rounded life capabilities including self-confidence, communication, and leadership. We have developed the following interactive Tableau view to describe FIRST:
Dow Jones Historical Trend: A Visual Analysis
With the recent drop in the Dow Jones, I wondered about the historical trend. I created a workbook with monthly average data I found on the web. Note that my expertise is in visual analysis, not finance, and this analysis comes from that perspective.
With a semi-log plot you can clearly see a generally linear trend to the data:
In Tableau, I can fit a linear trend line to this data by writing a calculation to convert the month to log(month, 10). As you can clearly see, the partial average for this March is currently touching the trend:
You can also see that the Great Depression dipped significantly below the historical trend before the Dow returned to the trend. That is also true for other smaller movements around the historical trend line. If you took these dips to be a precedent for our current market, my visual guess for the bottom of our current correction would be next September around 3,500 for the Dow Jones:
It might be lower since we were above the trend for a long time. Converting back to the a standard trendline, we get the following view:
Of course, all of this assumes that the political and economic situation today is similar to that of the Great Depression, and that our financial system is similarly architected. All of which could be questioned.
City of Charlotte Wows Us with Innovative "Business Analysis Olympiad"
I just returned from one of the most interesting and innovative business events I've ever attended. The City of Charlotte sponsored a "Business Analysis Olympiad" to promote the business value of visual data analysis software as well as to create a community of visual analysts within the city’s key businesses. When employees use data and information more effectively, they make better business decisions and thus serve the citizens of Charlotte better.
The contest was created by the Business Systems Support group in the IT department and attracted teams from across the city’s 14 departments to learn about the new ways that they could visualize and analyze data. In other words, IT functioned as a pro-active consultant to business units to improve their present practices.
The data set for the contest involved the sinking of the Titanic, the origin of the passengers and their fares. 12 teams of 2 persons each used Tableau to analyze this data set and present their findings. These teams included the Department of Transportation, Charlotte Area Transit System, Charlotte Fire Department (they had fans that came to support them), Economic Development Office, Engineering & Property Management, Planning, and Solid Waste Services.
After working with Tableau and the Titanic data for about 1 week, each team had exactly 5 minutes to present their findings. All teams did a great job and many of the teams had impressive presentations given the limited time they had with the product and data set. We saw sheet linking, business dashboards, scatter charts, stacked bars, pyramids, various time series, integrated mapping, imported moving visuals across the ocean, summary tables, pie charts, tabular, annotations, good use of color, data labels, trend lines, bar in a bar, Gantt charts and table calculations. The city is a huge user of GIS and many of the teams took to our mapping capabilities in a NY minute.
Professor Robert Kosara of the UNC at Charlotte, Julie Burch, Assistant City Manager, and I were the judges. Choosing the winners was a challenge but after a detailed discussion we made the following choices:
- First Place -- "Trash Talkers" from Solid Waste Services. The analysts were Kimberly Jenkins and James Gray with Michelle Moore as their sponsor.
- Second Place -- "Research Methods" from Planning. The analysts were Ruchi Agarwai and Evan Lowry with Steve Patterson as their sponsor.
- Third Place -- "Quality CATS" from Charlotte Area Transit System. Analysts were Celia Gray and Shelly McKee with Cilia Gray as their sponsor.
The first place team, Trash Talkers, did an excellent job of using Tableau to tell stories with the data. They showed that there was significant empty space in some of the lifeboats. They also used an overlapping bar chart to clearly show that it was better to be a first or second class female than a third class male. Below is one of their visualizations. Also, you can download their packaged workbook (which can be viewed with either Tableau Desktop or free Tableau Reader).
Second place went to Research Methods. Here is one of their dashboards. Also, you can download their packaged workbook part A or part B (both of which can be viewed with either Tableau Desktop or free Tableau Reader).
Third place went to Quality CATS. Here is one of their dashboards. Also, you can download their packaged workbook (which can be viewed with either Tableau Desktop or free Tableau Reader).
Congratulations not only to the winners and the participants but also to Jim Raper and his Business Systems Support group for putting on such a creative, fun and educational data visualization event. With this kind of thoughtfulness, creativity and innovation, it's no wonder that City of Charlotte has been named a best place to live numerous times.
Showing Electoral College Impact With Overlapped Bars
You can see that despite the attention that the 2000 election brought to the electoral college, George W. Bush’s win was the 14th time that the ultimate winner earned less than 50% of the popular vote but more than 50% of the electoral college. Among the other thirteen times, 3rd-party candidates may have been the "spoilers" (as in Bill Clinton’s 1992 win over George H.W. Bush and Ross Perot). Nonetheless, it’s an interesting observation to realize how often this phenomenon has occurred.
Three interesting notes: John Adams (1796), Thomas Jefferson (1800) and John Q. Adams (1824) all earned a greater percentage of the popular vote than they did of the electoral college vote, with John Q. Adams not even hitting 50% of either. That was quite an election since it was eventually decided by the House of Representatives; for more details, see http://en.wikipedia.org/wiki/United_States_presidential_election,_1824.
Here is how you build this overlapping bar view of two measures in Tableau:
- Build a bar view
- Drag the second measure to the axis to combine it with the first measure
- Move the Measure Names field to the color shelf
- Turn off stacking
- Adjust the overlap order by dragging items on the color shelf
- Finally, write a calculated field to create an extra column to show when the electoral % was less than the popular %. Add it to the Measure Names/Values card.
More can be simpler when telling data stories
The Junk Charts blog had a posting about the importance of making a data view “as simple as possible but no simpler”, which used a great example from Professor Gelman’s recent book “Red State, Blue State, Rich State, Poor State”. His scatter view does clearly show the interaction of social and economic views. It is also relatively easy to see the social clustering. However, it is harder to see the economic cluster. I think you can add more to Gelman's data view to tell his story more effectively.
The following dashboard includes Gelman's scatter comparison of social and economic issues. It also contains two additional views integrated with the scatter. The social view on the right clearly shows the clustering that differentiates the red and blue states, with the battleground states between. In contrast, the economic view at the top shows a different clustering based on the economic status of the voters. In this case, adding more makes it simpler to tell the data story. (click here to download the packaged workbook)
Another way to add more to this data visualization is to back it with the raw data so that I could add to the story by creating new views. In particular, this view causes me to wonder what would happen if there is a shift in the relative importance of social and economic issues. Are there red and blue states that might turn into battleground states because they contain a large percentage of poor voters?
Life Expectancy in OECD Countries
The Junk Charts blog had a posting about web publishing that included a comment by reader 'DavidS'. He suggested using quartiles in a chart about life expectancy in OECD countries. Since the data was easy to download, I explored his suggestion in Tableau. Although the original chart is a good summary of the upward trend of life expectancy, David is correct that a more statistical view showing outliers demonstrates that the variance has increased even though the range has reduced.
Since I was creating a statistical view, I decided to use a density plot to show the distributions by year. This also allows you to see when data was not collected for a given country.
And since there were more countries than an effective nominal color pallet, I used distinct color hues only for the low outlier countries. In Tableau, you can use the highlight feature of the color legend to explore the data for individual countries. The static view clearly shows that some outliers have moved significantly toward the mean.
As David predicted, the quartile markings (i.e., the small black lines) all show that the variance has increased even though the range has reduced.
You can explore further with the free Tableau Reader application (details / direct download) and the attached Tableau Packaged Workbook, which also includes trend line and map views of this data. JunkChart's original blog entry is available here and you can find David's comment here.
Building Cycle Charts to Look at Trend Data
The January 2008 newsletter from Perceptual Edge is an excellent description of Cycle Plots by Naomi B Robbins Ph.D, which are a less known way to look at trend data. On her website, you can find a link that describes how you can create these views in Excel and provided the input data. I think it is great when people give you access to the data that is shown in examples. I used the same data to build a cycle chart in Tableau in a couple of minutes.
Here is the packaged workbook, suitable to use with both Tableau Desktop and Tableau Reader (our no-charge application for reading and interacting with visual analytics). The workbook also includes the two other trends she talks about in her articles. All three views are useful and Tableau makes it easy to switch between them and get many other views as well.
Comparing Standardized Math Tests: Check Out Your WA School District
Nevertheless, my interest in Washington State math instruction has been useful to my work at Tableau. As the Director of Visual Analysis, I need to use our products authentically to see and understand data so that I can know how to improve the user experience for our customers. Since I have an authentic interest in the math testing data about Washington State, I explore it as a way to identify where our product is easy and hard to use. The rest of this posting describes how I did this and what I learned about Tableau and the test data. At the end, I include a link to the workbook, which you can view in our free Tableau Reader if you are interested in looking at the data for school districts in Washington.
DATA:
The first step was to access the test data. Since I was interested in the WASL test (Washington Assessment of Student Learning), Google quickly lead me to the website hosted by the Offices of Superintendent of Public Instruction (OSPI) to report WASL data:
http://reportcard.ospi.k12.wa.us/summary.aspx?year=2006-07
I found it pretty easy to use this web interface to see data about the state or individual school districts but I found it inadequate for answering comparative questions such as the relative trends of two school district. What I needed were views of data that were designed to support comparison so I turned to Tableau Desktop, which is a great tool for designing interactive graphical views of data. To use Tableau Desktop, I needed the underlying data rather than the summary data that was reported in the web interface. Turned out the OSPI provides this data as a bunch of Excel files on the following page:
http://reportcard.ospi.k12.wa.us/DataDownload.aspxThis was great except that each file had a unique set of columns and different years used different names for the school districts. I quickly found this out when I loaded the files into Tableau and spent a couple of evenings cleaning up the files and joining the resulting data with an address file that OSPI also provided. The result is quite useful but much more work than access to a database hosted by OSPI.
VISUAL ANALYSIS:
Since I was interested in comparative questions, I got very excited when I saw that the download page included the test results of the ITBS, which is a nationally-normed test produced by the University of Iowa. It turns out that the WASL and the ITBS were given to most of the students in the state from 2000 to 2005, which means I could create a view that validly compared the two tests. However, I was interested in the data at both the state level and district level. Although Tableau does a great job of aggregating data to different levels of detail, the ITBS data was reported as national percentile ranking. Therefore, I had to be careful and not sum or average this data. The result was that I ended up with four sets of data (WASL, ITBS at the state and district level of detail), which resulted in the following view:
The top pair of views compares the WASL and ITBS at the state level and the bottom pair at the district level. The WASL data is percent who met standard and the ITBS data is the national percentile ranking. The lines starting at zero show the delta from the first year shown. The visual finding is pretty interesting, which is that the WASL tended upward while the ITBS stayed flat. Since the ITBS is nationally-normed, this view told me that I should not use the WASL results to inform my opinion about the math instruction in Washington State.
Many other comparative views can be designed for this data. For example, here is a view for comparing the WASL results of school districts:
This screen shot includes a quick filter on the right that can be used to select the school districts for comparison. You can download the packaged workbook and play with this view in the free Tableau Reader.
CONCLUSION:
This example contains some useful lessons:
- Data found on the web will probably require cleaning
- Be careful what you aggregate. For example, national percentile data should not be summed or averaged.
- Visual analysis is a great way to answer comparative questions as long as you have the freedom to design the appropriate view for your question.
RSS - Jock Mackinlay's blog 
