Introduction
In this article I would like to analyse the data of the students in my own department, Statistics and Insurance Science in Piraeus. The most recent study guide includes data about those who survive (graduate) from the school, such as the mean graduation grade and the years it took them to get there. To begin with, we notice that the data we are interested in is in a relatively organised form, in tables (which is good), although we don’t have it as a data file, it is part of a pdf document (which is not good). Fortunately, among the chaos of more than 20,000 packages there is also the {tabulizer} package, which solves exactly this kind of problem since it can read the tables inside pdf files.
This article is a little old. When I first wrote it, the package was called {tabulizer}. A few years later, when I tried to make some changes, I noticed an error in this particular article. The package has since been renamed and you can now find it as {tabulapdf}. So keep an eye out for it, as various sources online still refer to it by its old name.
Loading the libraries
As explained above, since I need to extract data, and in particular tables, from a pdf file, the tabulizer package (now {tabulapdf}) is essential. Unfortunately, while trying to install it I kept receiving an error message similar to this one. The problem seemed to be related to the {rJava} package, and to resolve it I followed the notes in this comment. After installing {rJava}, I managed to install the package.
Extracting the data
The study guide starts by giving some information about the university, continues with the basic details of the school, the teaching staff, the requirements for obtaining a degree, and goes on with a more detailed description of every course in the department. At the end of our department’s study guide it has become customary to publish some basic statistics about the students. We will work with the recently posted 2022 study guide, which contains data on admissions to the department, on graduates, on the structure of the student body and other data going back to 2004. As is evident, we are only interested in the last pages of a lengthy study guide, since it consists of 200 pages. We will start by downloading the guide from the given link and then, with the help of the pdftools package, I will select the last pages, from page 186 to page 190, so that I can later extract only the necessary tables.
This restriction is quite important so that I reduce the search space for the tables, which gives me a much faster execution of the command (5 pages instead of 200). The tabulapdf package then takes over, and from it we will need the extract_tables command, which collects all the tables from a pdf and gathers them into a list. In our case five tables were detected, so I received as a result a list containing five data.frames, data tables. It’s astonishing that within a few seconds I received all the information in a form ready for analysis. If I had tried to enter the data in more traditional ways, such as copy-paste, I would have spent over an hour!
Admitted students
The first table I received concerns admitted students, that is, those who manage to enrol in the school. There are three basic ways to enrol in a school. The first and most common way is through the National Exams, where you compete against all the students in the final year of high school for entry into higher education. The second way is available if you already hold a degree: you have the right, instead of going through the ordeal of the National Exams, to sit exams at the school you are interested in. Usually these exams are organised by the school itself rather than by the state (as with the national exams), and in most schools they are scheduled for December. A third option is to obtain a transfer. If you have been admitted to a department for which there is a similar one in another city, you can request a transfer to another department; however, the places are limited and the criteria very strict. If you want to read the list of equivalences between schools you can click here. For example, in our department (Statistics, Piraeus) students from the Statistics department of the University of the Aegean in Samos used to come and attend courses, hoping that they would get a transfer to our department, which is near Athens. In total, the Ministry of Education has designated 4 departments in the field of Statistics as equivalent, and as far as I know there are no others in Greece (at the undergraduate level at least).
Having summarised the ways of entering higher-education departments, let’s look at what applies in our own department based on the published figures.
In our school the total number of admitted students from all categories ranged between 200 and 250 up to the 2012-2013 academic year. Over the next two years it broke the barrier of 250 admissions. The situation then returned to the aforementioned range until 2019, when we had a historical record of admissions to the department with 292 admitted students. An important observation I made is how similar the shape of the line is, and how closely the curve of total students matches that of transfers, in essence, transfers are the unpredictable factor behind any strong fluctuation in admissions.
Having studied the absolute values of the admitted students, the number of students over the last 20 years fluctuates between 200 and 300, which makes comparison across years difficult. So I will also compute the percentage share of each category per year. A very interesting fact is that in some academic years the number of students who entered the department outside the National Exams approached one third of the total enrolments. This seems to have happened in two academic years, 2010-2011 and 2013-2014.
Student population
A demographic indicator that can give us a rough sense of the department’s situation is certainly its number of students. This is very simple if we consider the implications of a fairly large number of people to be educated with given resources. The overwhelming bulk of students are undergraduates, at a share of 95%. We could say the proportion is reasonable, and we expected undergraduates to dominate, since their study cycle is longer than that of postgraduates. Moreover, the department offers only two (fee-paying) postgraduate programmes, which makes a smoother flow into higher-level studies harder relative to the number of students.
The element that deserves particular mention is not the ratio but the absolute number of undergraduate-level students. Unfortunately, in recent years the indicator keeps rising uncontrollably, having surpassed 3,000 students! The only year their number fell was during the exam period of the COVID-19 pandemic, which was conducted online.
Of course, the chart is not clear enough for the students outside the undergraduate level, since their share is comparatively much smaller. So I will isolate the postgraduate and doctoral students.
Distribution of graduation grades
All of the above figures were in some sense demographic and did not carry any particular value. So let’s now examine the first interesting element, which is the distribution of graduation grades. In the figures published by the department, the grades have been split into four categories according to the grade at completion of studies:
- From the pass mark (5) up to 6
- From 6 to 7
- From 7 up to 8.5, and
- those who scored above 8.5
The figures reveal a particularly unbalanced distribution of grades. First of all, we notice that the share of people graduating with “Excellent” is particularly small, barely visible on the chart. The share of graduates with distinction ranges between 4.8% of the total graduates of the corresponding academic year and 0%, since there were years without any top-scoring graduates. Another problematic element is graduating with the lowest possible grades (the grade category from 5 to 6), since historically we have had such graduations for more than one third of total graduates. Undoubtedly, all of the above is somewhat expected given the difficulty of the school and the subject. What possibly points to problems within the school itself is when we split the grades into two broad categories: one group being the students who graduated with a grade above 7, and the other those who graduated with a grade below 7. There, over a series of years, we observe that the graduates of the second group make up 80%, and there were recent years when they made up 90% of them. Indicatively, the academic years 2009-2010, 2012-2013, 2016-2017 & 2017-2018 are among those with the worst performance.
Distribution of graduation year
One of the most disappointing indicators for our school is how many years it takes someone to graduate. This indicator is now important for many reasons. To begin with, Greece has recently voted for the removal of students and there is now a limit on the length of studies, which is calculated as follows:
where is the number of years the studies last. In most schools the studies last four years, with the exception of engineering and medical schools where studies last 5 years. By extension, based on the above rule, the study limit for our department is 6 years. But reality is quite a bit harsher. This particular school is one of the departments with a historically high share of students who have exceeded n+2 as a proportion of its total undergraduates, with recent ELSTAT figures ranking the Statistics and Insurance Science department of Piraeus 35th on this indicator.
The relationship between years and grade average
The distributions of grades as well as of years to completion are quite informative for grasping the poor situation in the department. To finish, I would like to present a dual chart showing the mean study duration and the mean graduation grade. This will give us a picture of where the average graduate of the department has historically stood. Historically, there is a remarkable stability in the degree grade, which has a particularly small deviation from 2009 to 2020, ranging between 6.3 and 6.43 for an entire decade. On the other hand, we have once again the most depressing indicator, that of the mean study duration. From 2014 onwards we have a continuous rise in the indicator, and within 4 academic years it increased by 1.6 years and remained at those levels. By now, the mean duration not only diverges from n+2 but converges towards 2n, that is, students in the department needed twice the predetermined study duration (4 years) to obtain their degree. It is worth noting that over the entire time horizon in which the indicator is recorded, it never fell below the limit.
Statistics vs Statistics
Our country has a total of four departments from which one can graduate as a statistician. In the previous sections I analysed the graduate figures for the Piraeus department, since it is the only one for which I found detailed data. On the other hand, I would like to extend this analysis to a comparison of the similar departments. The ideal benchmark would be the mean time to graduation per graduation year, however such figures are not available for every department. For this reason we will use ELSTAT data, and more specifically the figures on the distribution of years of study among active undergraduate students. This will yield two quantities: the number of students who have not exceeded n+2 years, and the rest. The comparison of departments will be made on the basis of the share of students above n+2 years relative to the total undergraduates of a department.
| Rank | University | Department | % students over n+2 |
|---|---|---|---|
| 29 | University of Piraeus | Statistics and Insurance Science (Piraeus) | 75.81 |
| 104 | Athens University of Economics and Business | Statistics (Athens) | 61.08 |
| 154 | University of the Aegean | Statistics and Actuarial-Financial Mathematics (Samos) | 54.17 |
| 304 | University of Western Macedonia | Statistics and Insurance Science (Grevena) | 24.82 |
In this case we observe that the Piraeus department is by far the department with the most “eternal” students as a share of the undergraduate population, with three out of four being in their 7th year or above. This performance is not merely poor at the national level (29th), but considerably worse than the other similar departments. The Athens department has a share 15 percentage points lower (61%), followed by the Aegean department with one in two having exceeded n+2 years. The last Statistics department is that of Western Macedonia, and its results need some clarification. The department was founded relatively recently (2019), so it has not yet had time to “produce” students of higher years.
Statistics vs the other fields
Ok. We get it, if we study at UniPi things get grim (bordering on hopeless). But how do these departments really compare with the rest of the schools or fields? Which field ultimately gathers the most higher-year students in our country, and what could that mean?
| # | Category | Departments | Median | Min | Q1 | Q3 | Max |
|---|---|---|---|---|---|---|---|
| 1 | Geology & Geosciences | 7 | 76.2 | 65.8 | 73.3 | 78.9 | 85 |
| 2 | Mathematics | 9 | 75.3 | 23.4 | 59.3 | 80 | 84.9 |
| 3 | Physics | 11 | 74.1 | 22.5 | 45.6 | 75.2 | 79 |
| 4 | History & Archaeology | 13 | 67.9 | 40.2 | 62.9 | 71.2 | 74.5 |
| 5 | Philosophy & Theology | 10 | 65.9 | 23.6 | 52.2 | 82.3 | 89.3 |
| 6 | Philology | 18 | 64.1 | 43.7 | 55.7 | 76.5 | 90.5 |
| 7 | Fine Arts | 8 | 61.5 | 38.2 | 54.3 | 64.6 | 67.4 |
| 8 | Sports Science | 1 | 60.1 | 60.1 | 60.1 | 60.1 | 60.1 |
| 9 | Theatre & Cinema | 7 | 59 | 15.4 | 35.4 | 64.7 | 74 |
| 10 | Computer Engineering | 6 | 57.8 | 1 | 18.9 | 59.7 | 66.7 |
| 11 | Statistics | 4 | 57.6 | 24.8 | 46.8 | 64.8 | 75.8 |
| 12 | Naval Architects | 1 | 56.1 | 56.1 | 56.1 | 56.1 | 56.1 |
| 13 | Law | 17 | 53.5 | 0 | 22.2 | 62.1 | 75.9 |
| 14 | Other Engineers | 23 | 52.7 | 0 | 29.6 | 59.1 | 72 |
| 15 | Computer Science | 18 | 50.7 | 7 | 17.7 | 60.9 | 100 |
| 16 | Chemistry | 6 | 50.7 | 42.7 | 45.4 | 54.9 | 70.4 |
| 17 | Public Administration & Communication | 11 | 50.5 | 18.5 | 25.9 | 52.4 | 77.8 |
| 18 | Electrical & Computer Engineering | 11 | 50.3 | 0 | 20.6 | 56.2 | 81.5 |
| 19 | Economics | 15 | 50.2 | 0 | 45.6 | 63.4 | 80 |
| 20 | Foreign Languages & Translation | 3 | 49.9 | 20.4 | 35.1 | 61 | 72.1 |
| 21 | Chemical Engineers | 5 | 49.2 | 0 | 43.6 | 52.3 | 57.3 |
| 22 | Civil Engineers | 8 | 47.2 | 0 | 35.6 | 60.3 | 68.2 |
| 23 | Social & Political Sciences | 19 | 45.8 | 13.4 | 33.2 | 54.6 | 70.6 |
| 24 | Mechanical Engineers | 11 | 45.7 | 0 | 25.2 | 61.4 | 75.5 |
| 25 | Architects | 9 | 45.4 | 0.2 | 42 | 49 | 55.8 |
| 26 | Music | 6 | 45.4 | 0 | 11.9 | 56.1 | 63.2 |
| 27 | Biology & Biosciences | 10 | 45.3 | 0 | 39.5 | 58.4 | 74.2 |
| 28 | Shipping | 3 | 42.9 | 7.5 | 25.2 | 47.6 | 52.3 |
| 29 | Pharmacy | 3 | 41.7 | 41.1 | 41.4 | 50.7 | 59.6 |
| 30 | Business Administration | 15 | 40.5 | 0 | 21.9 | 64 | 80.8 |
| 31 | Psychology | 6 | 38.9 | 7 | 33.2 | 44.8 | 50.9 |
| 32 | Education | 22 | 35 | 8.1 | 22.7 | 40.3 | 54 |
| 33 | Medicine | 14 | 30.3 | 0 | 16.6 | 35.7 | 42.7 |
| 34 | Nursing & Allied Health | 28 | 24.9 | 9.9 | 21.6 | 45.2 | 61.5 |
| 35 | Accounting & Finance | 2 | 23.9 | 18.5 | 21.2 | 26.6 | 29.3 |
| 36 | Food Science & Nutrition | 14 | 20.4 | 0 | 11.7 | 41.4 | 68.3 |
| 37 | Tourism | 3 | 19.2 | 0 | 9.6 | 36.5 | 53.7 |
| 38 | Agriculture & Environment | 25 | 10.2 | 0 | 0.8 | 49.3 | 68.1 |
Of course, I would like to warn against any mistaken or hasty interpretations of the table above. The fact that a field has a high number of higher-year students is not proof of a department’s difficulty. It can, of course, be an important indication. The problem with this hasty link, “share of eternal students → department difficulty”, lies in the fact that studies in Greece have a more flexible character, allowing someone to stop their studies and return to them later. The reasons why someone may delay obtaining a degree can be many, and beyond the difficulty of the school. Perhaps the very subject of the school constitutes an important reason for the creation of a particularly high number of deferred graduations. For example, students may perceive that their department is not interesting or does not have significant career prospects, and thus have a low motivation to study and consequently to graduate. Obviously, there are also other reasons why someone may defer their studies, such as family reasons, health, etc. However, there is no particular reason that these factors affect one school to a greater degree, or at least I do not have such data to check it.