Reading and writing and … location. Visualizing where different peformance metrics correlate.
5 February 2010
Our parent company, Universal Mind, was tasked by the Colorado Department of Education and Center for Assessment to visualize data from their innovative models for measuring student progress. The public version of that project is available at schoolview.org. SchoolVIEW has some great features to visually compare school performance in terms of proficiency and growth (improvement over prior years) in reading, writing, and math. (You can learn more about the project here.)
CDE's SchoolVIEW
SchoolVIEW data in SpatialKey
I was interested in seeing the correlation between these different metrics, and (since we’re obsessed with location) how that correlation relates to geography. So, I imported that data into SpatialKey. The source file was a CSV with a row for each school. Here’s what that data looks like:
SpatialKey’s bivariate renderer allowed me to quickly explore the data in just that manner. The bivariate renderer allows you to select two numeric attributes in your dataset, and an aggregate calculation for each. In the image below, I selected average math growth percentile and average math proficiency. Each dot in the scatterplot legend at the upper right represents a colored location (grid cell) on the map. The position of the dot represents its relative score for average math growth (y axis) and average math proficiency (x axis). The color “behind” the each dot is the color used for the corresponding grid cell on the map.
This visualization shows the coorelation between math proficiency and growth, as it relates to location. (Click the image for a larger view.)
We can see there is a general positive correlation, where most locations have a similar relative score for math performance and growth: Most points on the scatterplot are along an imaginary diagonal line from the bottom left (low in both metrics) to upper right (high in both metrics). What’s often interesting and informative is to see areas that deviate from the norm in terms of the correlation. Areas with relatively high proficiency but low growth – “strong but losing ground” – are colored blue, while areas with low proficiency but high growth – “risin’ up” – are colored red. These are both negative correlations. Areas that score low on both metrics are shaded white, while those high in both attributes are shaded black – both positive correlations. For this type of visual analysis, areas that fall toward the middle of both ranges are usually less interesting, and so those colors are more transparent to allow you to focus on the extremes. It may take a few seconds to orient yourself to this view, but once acclimated it’s a powerful way to visualize some complex – and otherwise difficult to express – relationships.
You can correlate any pair of attributes by simply selecting from one of the axes in the scatterplot legend. This next image compares average math and reading proficiency. First, notice there seems to be an even stronger correlation between these two variables than the previous set. (The points line up even closer on the imaginary diagonal line.) It’s also interesting to compare these two images; Notice how the schools in some locations are relatively strong (shaded black) or weak (shaded white) in both visualizations, while others show a particular weakness in one of the metrics.
Selecting a point on the scatterplot shows the corresponding location on the map. In this case, we've highlighted a school that is an outlier because it's relatively strong in math versus its perforrmance in reading, realtive to other schools. We can easily see this school is in Moffat County. (Click the image for a larger view.)
SpatialKey makes it easy to uncover and visualize these relationships, and to share them with others. From uploading the spreadsheet with school data to presentation, this only took a few minutes to create – without any programming or hassle. And, this is just the start. By adding filters we can see these trends for schools of certain sizes or types, or compare these trends over time.
Further Analysis
An interesting next step would be to see if there is any correlation between the areas that deviate from the norm school performance and property value changes. For example, are the “rising up” areas ones where real estate values have been growing faster than average, or gentrification is taking place. (Of course, determining causality is a whole different conversation!) One could bring additional real estate or demographic data into SpatialKey to help answer those questions. SpatialKey makes it easier to understand the relationships between disparate datasets.
Try it out for yourself
Don’t take our word for it. You can start uploading your own data and visually correlating it right away by signing up for the 30-day trial of SpatialKey. Or, contact us and we’ll be happy to walk you through the process.
No Comments Yet