Editor's Note:This blog post first appeared in August 2011, at the end of blogger Jennifer Baulier's summer internship. We are republishing her blog post now that JMP 10 has been released and the Micromaps add-in discussed in the post is now available in the JMP File Exchange. JMP 10 is the minimum version required to create Micromaps using this add-in. Download of the add-in requires a free SAS profile.
My name is Jennifer Baulier, and I am a summer student at JMP. I attend college at WPI (Worcester Polytechnic Institute), and applied for my internship at SAS through the R3 program. I am going into my sophomore year and am studying both computer science and video game development. I have really loved my time here and have learned a lot especially while working on my Micromap project, which is explained below.
In June, Xan Gregg blogged about a Micromaps presentation by Dan Carr. At the time, Graph Builder could approximate Conditional Micromaps minus a few key features. My project adds on to what Graph Builder was already doing in order to recreate several types of Micromaps (Comparative, Conditional, and Conditional with Percent Difference).
The main point of Conditional Micromaps is to allow the user to look at trends between regions and three additional variables set to X, Y, and Color. The image above is an example report. My Conditional Micromaps are different from what Graph Builder already does because 1) the user chooses three ranges for the X, Y and Color groupings, and 2) the user can modify ranges on the fly with slider bars. In my Micromaps, the color groupings always increase from blue, to grey, to red. In the example shown above (where X is expenditures, Y is student/faculty ratio, and color is ACT scores), Texas has low expenditures, a low student/faculty ratio, and medium ACT scores. A few trends in the image above:
All of the red ACT scores occurred in southeastern states.
All the blue ACT scores occurred in northern states
All red ACT score occurred in state with at least mid-range expenditures.
Unlike Conditional Micromaps, Comparative Micromaps are not very close to anything that already exists in Graph Builder. The center panel is a set of maps organized by an X grouping and split into three color ranges chosen by the user, which can be updated with the slider. The top and bottom panels show changes from one color to another between groupings. The top panel shows increases while the bottom shows decreases. In the example, the states shown are colored by average unemployment rate across a set of years. If you look at Florida, you can see that in 2007 it was blue, but in 2008 it increased to gray. Since it changed to a higher category, Florida is colored gray in the top panel map for 2007 to 2008. Several major trends appear:
2007 to 2009 showed a huge increase in unemployment.
2010 to 2011 showed improvement.
Most states that hit the red zone were in the South.
Three consecutive central states stayed blue the entire time.
Comparative with Percent Difference
Comparative with Percent Difference (shortened to “Percent Difference” later in this description) is a variation on Comparative Micromaps that I implemented. Like Comparative, the center panel shows the data over some X grouping, and the top and bottom show increases and decreases. Instead of displaying the color in terms of three ranges, the color is displayed in a continuous gradient. Also, instead of the top and bottom panels showing increases or decreases between color ranges, they show increases or decreases for which the change is at least N percent different. N is chosen by the user and can be interactively changed with the slider. The image above shows Triangle (North Carolina) real estate values over time. The shapes shown on the map are based on ZIP codes (I created custom shape files using data from the US Census Bureau). The slider is set to a fairly low percent so that most changes will show up on the map. Note the following:
The first two boxes on the decrease panel reflect the dot-com crash in 2000-2001.
The next six boxes show the economic boom that followed.
Chapel Hill, North Carolina, (the area on the far left) goes from light blue all the way up to bright red, showing how extreme the economic boom was for some areas.
The last two boxes in the decrease panel signify the current economic downturn.
Comparative Versus Percent Difference
Both types of maps are useful, but there are certain situations where using one opposed to the other would make a significant difference.
Comparative is most useful when:
When you have thresholds you need to observe.
When you want to give a simplified report that is easy to view.
Percent Difference is most useful when:
When you need to filter out a given amount of variation.
When you have a few outliers making setting ranges impractical.
When you want to find the most drastic changes.
When you intend on showing the duration of a period of increase or decrease.
Micromaps are a clear and concise way to view both regional trends and correlations in data. Furthermore, the interactive nature of the maps makes it easy to adjust the report in order to find helpful information. Comparative and Percent Difference Micromaps focus on deltas while Conditional Micromaps focus on correlations.
The slider bars in this project are not a standard display box item and were my design. You may see two-prong as a display box item in future versions of JMP as a result of this project.