The GPA may be severely biased. I am not an expert in GPA's history and details, but I believe I am an expert in measurement and assessment, especially in higher education., especially with respect to the well-known problem of how peer assessments may influence (=have an impact on) lecturer's score, and hence grade, of the members ("peers") of a team of students working on a group assignment.
Why do I suggest that GPA is severely biased? Because as far as I know, GPA (points) is based on Grade Points which belong to a grade scale which is bounded (=has lower and upper limits) and even finitely ordered (=contains only a small number of ordered points).
Since long, there is a ongoing discussion between experts and non-experts whether such scales may be treated as interval scales or not. But that is only half of the underlying problem, or only a certain perspective on it. The other half, or perspective, is how you may and should calculate a meaningful average (mean) on a strictly bounded range of possible values, irrespective of the issue which kind of scale is applicable or not (Stevens' statistical perspective).
I recently found out how to define that in a principled way, using the almost 90 years old concept of quasi-mean. It turns out, that there is a well-defined concept of mean for such bounded ranges which is clearly different from an arithmetical mean or geometric mean. Actually there are two such means depending upon whether the bounded range shall be standardized to the unit interval ranging from 0 to 1 (like percentages) OR to the interval ranging from -1 to +1 (like bipolar Likert scale points).
Conclusion: Be careful when using results of empirical studies which erroneously calculate the GPA as if an arithmetical or geometrical mean of grade points would make sense for such vales!
Thanks, Ette Etuk, for your prompt answer ;-) The correct GPA (grade point average = mean of grade points on a strictly bounded scale) that I have discovered is also based on the assumption that we are dealing here with a kind of mixed ordinal/interval scale. That's not my point.
Neither is your comment that it has been / is "a longstanding criterion for academic evaluation": people, even academics, may err for a long time, for whatever reason (examples abound).
The only advantage I see is, that it is used by many universities over the world, so there is a sort of tacit tradition or convention and agreement to use it so that no quarrels about it will arise. Still, it is based on an assumption, no hard facts: the assumption is that it is the best one can do.
However, it is a mathematical fact that one can do better than that. Even if in some cases the differences may appear to be negligible, GPA based on faulty premises, when used on a large scale, may indeed lead to quite different, probably embarrassing results. Only empirical research can tell us for sure. And it doesn't seem to be too complicated to conduct such research, given that the data on which GPA's or similar constructs rest are stored in many administrative databases over the world.
If there happens to be a well-documented research paper which tells me that and why the GPA is a perfectly valid measure of academic performance, better than any other conceivable measure or means, I will be highly interested to examine it and revise my current opinion. Until that moment, I prefer to stick to the hard facts that I discovered.
- https://en.wikipedia.org/wiki/Grading_(education)
- http://psycnet.apa.org/record/1976-08370-001
- Errors in GPA: Sources identified
Article Grades and Test Scores: Accounting for Observed Differences
- https://eric.ed.gov/?q=college+grade+point+average+
- https://eric.ed.gov/?q=grade+point+average+calculation&id=EJ992990
- http://gpacalculator.net/college-gpa-calculator/
@ Sergio López Bohle : Thanks for your response.
Dear Ian: many thanks, as usual. I have just "requested a copy" for the article which will probably be the most interesting and based on factual research. The wikipedia-article is not informative enough (for my purpose): it diesn't really tell what a grade is. Seemingly, it can be anything.
Paulus
Ian: From what I have read quickly in the other sources you mentioned, everybody seems to take the GPA for granted. There is some discussion and research about GPA-on-ordinal-scale versus GPA-on-interval-scale, but that doesn't even touch the deeper problem about bounded versus unbounded scales. The calculators I've seen all use few (integer) numbers although Wikipedia states that grades can be almost anything (e.g. alphabetic symbols, ranges of scores, what you like). The calculators don't even explain in clear terms what kind of scale they are using and which kind of mean (average) is being calculated. We have to dig deeper, e.g. by using very specific search term in Eric leading to less than 100 hits.
From this distance, I think that the problem is that
a) Teachers realize that scoring is not an exact science. They are happy giving what is effectively a Likert score.
b) This saves the teachers time because all they have to do is sort the essays into 5 piles.
c) Each student's marks for a subject for the year are _added up_ and used to determine subject mark and subject pass / fail for the year. At schools weighted averages do not seem to be used.
c) The best scores are used to decide subject awards at the end of the year. This seems OK, but other algorithms can be investigated.
d) All student's marks for the year for a subject are added up and used to determine class "average". This is published on the end of year report, together with the student's mark for the subject for comparison. The parent can see how well their child is doing compared with the others. This is useful if e.g., the marking standard was too high. and the child's mark is not representing properly how much the child has learned. The only other method is to grade on the/a curve. A simpler approach would be to give rank in class, but this seems not to be politically correct.
e) Researchers then use grade as a variable to find correlations with drunken behaviour, gender, ethnic group, employability, etc. generally proving what is common sense, anyway. Researchers know that the grades are only proxies for measuring knowledge and skill acquisition, but it is all they have got.
Assuming that Ian's description of the assessment practice at schools and universities is approximately representative for most countries, my first reaction would be, that teachers and researchers keep each other in a fatal bond:
- teachers believing that they can't do better than practices a) + b) + c) + d) (what is evidently not true);
- researchers taking what is currently easily available as a "convenience" sample e) (lowest level of empirical data with almost no chance to find meaningful correlations, not to speak of causal connections);
- teachers believing what comes out of current research most be somehow valid, implying there is no need to improve assessment practice, etc.
This cycle goes on and on, because both parties reinforce each other in their false beliefs.
I think I begin to understand why many people claim that real progress in the field of education is so extremely difficult to achieve. It is not just external factors like lack of money and time, it may well be internal factors like false beliefs and inappropriate methods.
The irony: teachers and researchers are in principle so close to the source of knowledge and skills which could catapult them in a new and fresh era of education, including educational assessment: their own alma mater, or at least other universities and research institutes specializing on those topics.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics