Every year, schools are required to have students sit for standardized tests. These tests cover math, quantitative reasoning, and English and Language Arts studies. One significant year for students, especially in the NYC area, is 7th grade. These scores help students apply to private and public schools for the next year. They also let parents know what high schools are reasonable for their students to apply to in the following cycle. These scores are recorded on a scale from Level 1 to Level 4. Level 1 and 2 are below grade level, and Level 3 is on grade level, with 4 being high proficiency/above grade level.  Level 1 means that students are significantly below grade level. In years past, these would be students who may have potentially been held back. 

Since schools are spread out over a region, I decided that to help myself better understand the data, it would be helpful to look at it spatially and organize it based on district, so that I would have more data entries for a larger region. I also chose to use a percentage rate – since it felt more “normalized.” I was concerned normalizing the data would skew it additionally, since some areas have less entries recorded.

Figure 1: 2015 – Percentages of 7th Grade NYC Students Scoring a Level 1 on the Math Regents

​

Figure 2: 2013 – Percentages of 7th Grade NYC Students Scoring a Level 1 on the Math Regents

​​After some self-reflection, some areas that I noted after making the maps for growth were adding in more layers for population, normalizing in a different way, learning how to make templates in ArcGIS to display the data better, and adjusting the decimal places shown & adding percentage signs.

I then interviewed two of my coworkers who are both NYC middle school teachers. They gave the following feedback:

​- Both preferred the color scheme of the first better than the many shades of orange. 

– The dots were confusing for them to understand.  

Phase two for me was changing colors and making the legend clearer. I realize now that the legend is inconsistent (some have more than one % sign). As a final adjustment, I would make the following changes:

– Adjust the gradients – experiment with color more. 

-Find a way to layer both over each other so that you could see the change over the two years and highlight areas that had growth/loss

-Edit the keys to better reflect and signal percentage data

-Remove the dots completely OR keep maybe a couple click-able significant schools

Figure 3: 2013 – REVISION Percentages of 7th Grade NYC Students Scoring a Level 1 on the Math Regents

Figure 4: 2015 – REVISION Percentages of 7th Grade NYC Students Scoring a Level 1 on the Math Regents

Both of these looked visually more appealing and gave more context to the data. Because of the strict level 1 focus, I think the map isn’t the most compelling option. I decided to shift my focus to make a more impactful map. 

One way that felt more visually straightforward was to organize it by average math score for the district and organizing the colors by level. 

Figure 5:  (Left) 2015-Average scores for 2015 Public Schooled 7th Graders in NYC, organized by district; (right) School Attendance by District for 2015

While this seemed more visually clear, it also pushed a narrative that isn’t necessarily supported by the data set. The difference between a high level 1 & a low level 2 is about 20 points. If a section of scores were left un-recorded, this can skew one district from being a “true” level 1 or level 2 or level 3. To fully understand why the map looks this way, a cluster analysis would have to be done & likely looking at the school as individual entities rather than districts would prove more helpful. The current map smooths – over and underemphasizes regions of change.

I then decided to compare both years to school age poverty in the same areas. The poverty map is organized by zip code instead of district, which isn’t ideal. However, I think it gives a fuller picture of why some data looks the way that it does. It makes the data on the left looks less misleading. I left one without a key– I think some type of continuous scale would best represent it, which I need to learn how to make on ArcGIS. 

Figure 6:  (Top Left) 2013-Average scores for Public Schooled 7th Graders in NYC, organized by district; (Bottom Left)2015-Average scores for Public Schooled 7th Graders in NYC, organized by district; School Age Population in Poverty (Right)

If I had more time, I could have made this an interactive map so they could interact with one another. I also could have spent more time with the population data & organized it by district so that they visually line up more fully. I also would have done some clustering analysis as I think organizing it by district smooths the data in a misleading way. 

Overall, my takeaway from maps would be that they can be a powerful tool if used well. I think some data (like this topic) lends itself well to maps, as usually there’s outside environmental and social factors beyond the data that need to be shown in conjunction. Maps, I think, are very difficult to show on their own in a way that’s not biased/misleading. Learning how to be thoughtful with the text around the maps should have as much impact as the map itself.