Data Exploration on 2015 World University Rankings
The skills the author demoed here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
Contributed by Ho Fai Wong. He is currently in the NYC Data Science Academy 12 week full-time Data Science Bootcamp program taking place between April 11th to July 1st, 2016. This post is based on his second class project - Shiny visualization (due on the 4th week of the program).
Several institutions score and rank universities across the world every year. What were the 'best' universities in 2015? How do these rankings differ by ranking organization? These were some of the questions that motivated me to develop a Shiny and R-based data application to explore the 2015 world university rankings provided on Kaggle by the following 3 ranking organizations:
- Academic Ranking of World Universities aka Shanghai Ranking
- Times Higher Education
- Center for World University Rankings
In my particular case, I was curious how the French "Grande Ecoles" performed in these international university rankings since I had studied in one. My initial perception was that these prestigious schools are very well-regarded in France, but relatively unknown overseas. Can the 2015 world university rankings from these 3 organizations confirm and provide additional insights to this assumption?
You can try out the application yourself here.
All the code is on GitHub here.
II. Preparing the Data
Kaggle provided the multi-year international university rankings in separate CSV files. I focused on the 2015 rankings, which covered a total of 1015 institutions across 61 countries:
- Shanghai Rankings: 500 universities
- Times World University Rankings: 401 universities
- Center for World University Rankings: 1000 universities
Considerable cleanup was required before building the Shiny application:
- University names: The only university identifiers in each data file were the university names, and these weren't consistent across the 3 ranking organizations. As a result, I reconciled university names across the 3 ranking organizations to ensure consistency
- Countries: The Shanghai data set did not contain country information, but Kaggle provided a separate mapping table of universities to countries to fill in this gap partially. I filled in any remaining gaps in country names, and reconciled country names across the 3 organizations
- Missingness: There were a few cases of missing criteria scores which I assumed were Missing At Random and imputed using K-Nearest Neighbors with k=sqrt(n) (e.g. 0.4% of the N&S criteria for Shanghai data and 5% of the income criteria for Times data
- World ranks: Shanghai and Times rankings stopped giving individual ranks to universities after the ranks of 100 and 200 respectively, and instead binned them into ranges. For visualization purposes, I assigned the middle of those ranges as the rank. Similarly, some universities had equal ranks (i.e. '=134' in the data) so I assigned the same rank to each of those institutions
- Criteria ranks to scores: Shanghai and Times data had scores (0 to 100) for the sub-criteria but CWUR data had ranks instead. For clarity and relative comparison purposes, I converted the CWUR sub-criteria ranks into scores ranging from 0 to 100: score = (1-rank/max)*100
III. Building the App
The Shiny application has 3 main sections:
- Country Statistics: world map and bar chart of key metrics by country
- Organization Comparison: university rankings compared between organizations
- University Profile: for a given university, its sub-criteria scores for each ranking organization
Country Data Statistics
The first tab of the application summarizes university ranking key metrics by country. The user can select which ranking organization's data to display, followed by which specific metric to represent as each country's color (e.g. top rank, median rank, etc). Since each organization used different scoring criteria, the list of metrics will change depending on the user's selection of the ranking organization (using a conditional panel). Definitions for these criteria can be found on the Reference tab.
The same data, filtered by the same selections on the left panel, is also displayed as bar charts with ordered countries.
It becomes apparent that, depending on the ranking organization selected and criteria considered, a country's universities could perform quite differently in these assessments. For example, while the United States retains the top-ranked university in all 3 rankings, Russia's top university dropped by 110 positions between the Shanghai and Times rankings.
Let's look closer at the comparison of rankings between these 3 organizations.
Organization Comparison Data
The second tab of the application compares university world rankings between organizations on a scatterplot which can be customized by the user to gain specific insights. The user can:
- Select which organizations to compare i.e. to display on the vertical and horizontal axis
- Select one or more countries to display on the scatterplot
- Zoom in/out on the scatterplot and hover on a point for additional information on the university
- Filter universities based on the specific metrics and criteria for each of the selected ranking organizations. Definitions for these criteria can be found on the Reference tab
In addition, the diagonal line represents the case where both organizations assigned the same rank to the university. Conversely, the farther a point is from the line, the greater the discrepancy between the organizations' assigned ranks for that university.
Just looking at the comparison of Shanghai against Times rankings, quite a few universities are positioned far from the diagonal line i.e. there is a large difference in how each organization ranked them. For example, Ecole Polytechnique (one of the most prestigious Grandes Ecoles in France), was ranked 61 by Times and 350 by the Shanghai ranking!
Using the sliders to filter on each organization's metrics can also lead to interesting observations. For example, according to the Times rankings, the United States of America seems to have very few universities with a high "international score" compared to other countries (only MIT in the USA has such a score greater than 75)...
Let's look at specific universities to compare how each ranking organization's criteria scoring coincide with or differ from each other.
The third tab of the application displays, for a selected university, the world rankings by each of the ranking organizations along with the scores for the underlying criteria. As described earlier, the CWUR data contained ranks for their criteria whereas the Shanghai and Times data contained scores (from 0 to 100). For consistency and ease of comparison between the ranking organizations, CWUR criteria rankings were converted to scores out of 100: score = (1-rank/max)*100.
A user can select a specific university to see and compare how each organization assessed that institution using their respective criteria. Keeping the example of Ecole Polytechnique, this university scored much lower in the Shanghai rankings than the Times rankings; the former places a much heavier emphasis on research and publications whereas the latter also considers international outlook and industry income.
Data and Reference
The Data tab allows a user to search for a specific country or university and obtain the world ranking data in tabular form for each of the 3 ranking organizations.
The reference tab contains background and comments on the data preparation which I mentioned at the beginning of this post. I also included here a graphical view of the difference in universities in scope across the 3 ranking organizations, as well as correlation plots for each of the ranking organizations' criteria.
Hopefully, this Shiny application can help you explore world university rankings from the Shanghai, Times and CWUR rankings: look up your own university, compare institutions for future students and see how rankings and criteria scores vary depending on which organization assessed the institution!
One of the main takeaways here is that, before making any definitive statements or decisions based on these rankings, one needs to understand precisely how each organization's criteria are defined and universities assessed against them. Standardizing such rankings is complex and can often be controversial:
- While the Times ranking is widely regarded as one of the most influential and widely observed university measures, it has been criticized for its commercialization and for undermining non-English-instructing institutions
- The Shanghai ranking, an equally influential ranking, has been criticized for focusing on raw research power and for undermining humanities and quality of instruction
Possible future steps for this project would be carrying out statistical analysis to investigate the impact and importance of each organization's criteria, and potentially designing a more balanced ranking framework.
In the meantime, please feel free to explore this application and glean more information on the 2015 world university rankings (and discrepancies between ranking organizations...). If you have any comments, suggestions or insights, please comment below!