Computer Science Educators Stack Exchange is a question and answer site for those involved in the field of teaching Computer Science. It only takes a minute to sign up.
Sign up to join this community
My university has decided that it should give a core (that is compulsory) statistics class to its computer science undergraduates. This opens the interesting question of what should be in such a class that everyone has to take.
As far as I know most CS degrees don't currently contain statistics classes so it's hard to compare to what already exists.
Does anyone have any experience of this and what would people recommend for a new statistics class for computer science students?
My first thought to provide a bit of something for you to go on was CS2013: The ACM/IEEE Joint Curriculum Guidelines for Undergraduate Degree Programs in Computer Science. This is available at https://www.acm.org/education/CS2013-final-report.pdf. I've copied out two relevant passages from that here:
"Computer science curricula should be designed to provide students with the flexibility to work across many disciplines. Computing is a broad field that connects to and draws from many disciplines, including mathematics, electrical engineering, psychology, statistics, fine arts, linguistics, and physical and life sciences. Computer Science students should develop the flexibility to work across disciplines." (p.20)
"Similarly, while we do note a growing trend in the use of probability and statistics in computing (reflected by the increased number of core hours on these topics in the Body of Knowledge) and believe that this trend is likely to continue in the future, we still believe it is not necessary for all CS programs to require a full course in probability theory for all majors. (p.50)".
It is worth searching that document for 'statistics'. It is mentioned in many other areas, but mostly as something learned in other areas such as Networks, HCI, Cryptography, etc.
The December 2013 issue of ACM Inroads had a good article (http://dl.acm.org/citation.cfm?id=2537777) on the role of mathematics in CS. This included a section on "The Current State of Mathematics in Computer Science Curricula". The article states that "The most-connected mathematical topic [to CS] by far is probability and statistics" and has some references to such. They also present the mathematics requirements of 25 'high quality' CS programs. Stats and Prob was required for 15, Calculus for 21 and Discrete Mathematics for 22.
http://dl.acm.org/citation.cfm?id=1240202 discusses statistics in liberal arts CS curricula.
Searching the CITIDEL syllabus collection (citidel.villanova.edu/ then look for syllabus collection) for 'statistics' might be worth a try, although the website has been varying between unresponsive to slow to responsive lately. Also, some of the material in there is a bit dated, but then again, statistics dates fairly well!
Given that in 2013 60% of sampled 'high quality' CS programmes had statistics, maybe it is more prevalent than you think. Perhaps the best way to obtain the most up to date information is to check the websites of several good CS programmes and hope that if they include statistics, the syllabi are available online.
I recently completed the MITx MOOC 6.00.2x, which is modeled after an introductory programming class in Python at MIT. The course addressed the following topics from statistics that integrate nicely with computer science (especially data science and machine learning):
Depending on the level of students and their familiarity with some basic concepts from statistics, I would also recommend looking at HackerRank's 10 Day of Statistics challenge. While the pedagogy isn't the greatest, it does scaffold nicely from mean, median, and mode through standard deviation and distributions to end up at linear regression and correlation coefficients. There is also some discussion of probability along the way.
Taken together, these two examples provide some big picture concepts to address that are fundamental to stats and that have relevance for computer science and programming.
I think that statistics courses are most useful for computer scientists who are anticipating a career in research. While statistics can be useful for professional developers that knowledge is not as critical as it is for researchers. The CS2013: The ACM/IEEE Joint Curriculum Guidelines for Undergraduate Degree Programs in Computer Science breaks knowledge units into three broad categories.
Statistics falls into the third category. I would not make statistics a required course but if you are going to do so I would suggest you look at the sort of topics you most want researchers to know. Ideally you are providing your students with the knowledge to properly evaluate research so that they can make informed decisions based on it. You'd probably want a deeper level for people actually doing research.
FWIW I was on the CS2013 task force but opinions here are my own and I am not speaking for the task force.
I got my CS degree in the 1990's and at the time (and perhaps still today) there was a lot of crossover between math and CS. By just taking the minimum math requirements to meet my CS degree I was one math class away from a math minor.
Subsequently in 25+ years as a software engineer I have never used most of that math, the only math I've actually used is geometry, trig, and probability/statistics. Especially these days with "big data" being a popular topic in computer science, and map/reduce being a primary technique, the concepts of frequency, concentration, correlation, regression, distribution, expected values, etc. are very applicable to problems that you see every day.
Also the type of thinking and reasoning you need for stats and probability is much more applicable to real-world problems since it has more to do with estimation, approximation and trends. One of the faults in my opinion with math is that you are often expected to arrive at a single clear and concise answer, when in fact few real world problems are solved that way.
Conversely I've never once used all of that calculus (which is a shame because I liked it) and related math. It was fun to take and probably a good (extended) mental exercise and weed-out for students but I think the stats and probability should have been emphasized a lot more than it was.
