통계분석, 머신러닝을 이용한 데이터 분석

Socio-Economic Status and Performance of Mathematics

13 Jul 2016

I visualized relationship of socio-economic status and performance of mathematics from Programme for International Student Assessment(PISA) dataset. The explanatory question is: What would be the average performance if all students had the OECD-average socio-economic status?

PISA measures performance only among 15-year-olds who are enrolled in education. In 2012, PISA covers 65 countries and economies,including all 34 OECD countries and 31 partner countries and economies. PISA also measures index of economic, social and cultural status, so called ESCS. It was created on the basis of the following variables: the highest levels of occupational status of the student’s parents, the highest level of education of the student’s parents, family wealth, etc.

According to the PISA report, we can see socio-economic status and performance in two aspects:

The strength of the relationship between performance and socio-economic status varies across countries, so it is hard to compare the performance between countries. On average across OECD countries, a more socio-economically advantaged student scores 39 points higher in mathematics – the equivalent of nearly one year of schooling – than a less-advantaged student.

In addition, The countries with strong relationship between performance and socio-economic status also tend to have big performance gap across socio-economic status. On the other hand, the countires with weak relationship between them are likely have small performance gap across socio-economic status.

To compare math performance, this visualization focuses on comparisons of different education systems based on the performance of students with similar socio-economic status.

Thus, we need to assume that all students had the same OECD-average socio-economic status, and then compare the math scores between countries. Most education systems perform similarly before and after accounting for socio-economic status. However, some ranks are changed considerably.

If socio-economic status were taken into account, rankings of mnay countries are increased. In particular, three countries that would climb more than 10 positions in their performance rankings.

If socio-economic status were taken into account, rankings of mnay countries are decreased. In particular, two countries that would descend more than 10 positions in their performance rankings.

I selected the data from “PISA 2012 Results: Excellence Through Equity”, which contains very interesting insights related to equity in education. Escpecially, I read chapter 2 and chapter 5.

Design

Chart types

First of all, I used choropleth map to address overall participating countries and economies in the dataset. With popup of each country, I presented total and sample population of PISA.

Second, I applied slopegraph, because it is a good way to visualize ranking changes when socio-economic status were taken into account. You can find the answer in the slopegraph. Ranking of observed math scores is on the left side, and ranking of adjusted math scores is on the right.

Visual encodings

I used colors to visualize participaing countires and economies on the map. In the slopegraph, color and hue are used to present continents differences, and ranking up/down.

Layouts

I used navigation tabs. Because this is martini-glass style of visualization: starting from exploratory visualization, viewers can clearly see that some countires perform better or worse given their ESCS as the navigation tabs progress.

Legends

I added legend in the map to show categories of participating countires. In the first slopegraph, legend is to emphasis specific continent. You can hover the legend to see it.

Feedback and Iteration

I revised my visualization for two times.

Here are sketches:

Here are details:

Feedbacks of first sketch

Source codes are under “iter1” folder.

feedback1

The biggest question is how do you account for socioeconomic status? Is what ever metric you are using concrete? Does the fact that Iceland lowers mean anything? Or is it just because they have a high socioeconomic status? If I am to believe in the metric am I to believe that Vietnam, Portugal and Turkey have good education systems buy poor access?

feedback2

I agree with Doug’s input: what does it mean to beat the socioeconomic circumstances? how did you measure performance of an education system? … But here you have some more input about the visualization itself. You say there are only three countries that climb more than 3 positions and it is hard to find those on the graph. Instead of coloring the lines according to continents you could color the lines according to increment in positions (example: green if they climb more than 5 positions, orange if between +5 and -5, red if less than -5, etc.). Also, is it important to visualize the continent? Is there any message you want to send such as “continent X countries would (on average) improve the most”? If this is not essential you could just drop it. If it is, you could color the words themselves according to continents, you could give the user the possibility to visualize 1 selected continent at a time (filter by continent), etc. Ok, hope it helps :) Other than this, very cool visualization!

feedback3

It would be great if the ESCS could be found in the visualization. And how does the adjustment change the reliability of the results, especially for countries that are far from the average ESCS? As I understand it, ESCS is mainly influenced by parents’ education. But it is much harder to improve an already high level of achievement than a low level. Has this been taken into account? Additionally, a visualization based on the ranking of the country obfuscates the actual change of the score, which is higher for Vietnam than for Turkey. Last, your choice of geographic regions is poor.

feedback4

Very good visualization! My only comments would be:

  1. It seems that the tooltips are positioned absolutely whereas the chart is positioned relatively. So they don’t always line up nicely.
  2. What “rank” is this? If the rank (e.g. 613 for Shanghai) is a PISA score, maybe just make that explicit (and maybe give some information on what this score is for the ignorant, like me ;)

feedback5

Very good first glance. Interesting topic. It made me want to ask 1) What’s the min-max range? 2) What’s the OECD global mean/median? 3) Are the score points just aligned or scaled accordingly? 4) What does “adjusted rank” mean? Does that mean for top socio-economic status group the score is much higher (higher variance - uneven society)? 5) What are the socio-economic status index of each country? 6) How about reading or other subjects?

feedback6

Great viz! Clear annotations, good choice of colors, fonts and sizes. Well explained. Two things: The annotation bubbles are somewhat misaligned to the diagram. What would you say about presenting the quantiles of the math score in a box plot, instead of just listing them in the left annotation bubble? Best regards and keep on with the good job!

Feedbacks of second sketch

Source codes are under “iter2” folder.

feedback1

I like how you’ve broken out the slope chart into three charts to show the overall, focus on the ones who do better, and focus on the ones who do worse. I was actually just reading about a similar tactic in this blog post: http://stats.blogoverflow.com/2011/12/andyw-says-small-multiples-are-the-most-underused-data-visualization/ The map is a little confusing. I’m not sure what OECD means, or really what I’m looking at on the map. I can mouse over countries to get the name, but nothing else pops out. I’m not sure what you’re trying to accomplish. Perhaps you could pair the map side-by-side with the slope charts so that when you mouse over a country, it highlights that country’s line on the chart and shows the data that goes with it?

feedback2

Very good visualisation. A bit unclear what the map is for. I am not sure if it is needed there at all.

Feedbacks of third sketch

Source codes are under “iter3” folder.

feedback1

The formatting is still inconsistent. Please comment functional sections of your code so that it is clear what is being done. Make sure that your visualization has a clear message.

Resources

comments powered by Disqus