Coursera, an education start-up co-founded by Stanford Professors of Computer Science Daphne Koller and Andrew Ng, made teachers and students around the world take notice this summer when it announced that 12 of some of the most respected American universities - Caltech, Princeton, and Penn to name a few - would join its effort to create college-level courses that could be taken by anyone for free online. In terms of institutional partnerships, this makes Coursera the biggest initiative among the current triad leading massively open online courses (MOOCs). The other leaders include Stanford Professor and Google Fellow Sebastian Thrun's start-up, Udacity, and the edX project lead by MIT Professor Anant Agarwal. The buy-in from its academic partners has helped Coursera become the free online university with the largest and broadest curriculum. Its catalogue lists 121 courses in 16 different subject areas, from Medicine to Information, Technology & Design. Though only 18 of the courses have gone online, this is 8 more than the number available at Udacity and 17 more than edX.
In seeming anticipation of the 2013 Year of Statistics, three of Coursera's September offerings are new classes about reasoning with data. Andrew Conway, Professor of Psychology at Princeton University, gets things underway with Statistics One. Later in the month, Professor of Biostatistics at Johns Hopkins University Roger D. Peng will offer Computing for Data Analysis and Professor Brian Caffo, who is also in the Biostatistics Department at JHU, will run a Mathematical Biostatistics Boot Camp. Statistics One is a six-week gentle introduction to statistical reasoning with no prerequisites. Brian Caffo's 7-week long 'boot camp' in biostatistics, as the name suggests, will be more technically rigorous, presenting material intended for first-year graduate students with some knowledge of calculus. Although all of the courses plan to include some programming with the statistical language R, learning and using R will be the focus of Roger Peng's 4-week Computing for Data Analysis.
At Coursera, course material is released on a weekly basis and typically includes a set of video lectures, quizzes, and homework assignments. Instructors anticipate students will usually need 3 to 5 hours per week to review material and complete assignments. Enrollees are encouraged to keep to the weekly schedule as best they can but only assignments have set deadlines. The door to the virtual classroom is otherwise always open. Students can go at their own pace and review material ad infinitum, like a favorite tune on the iPod.
In Conway's introductory video to Statistics One he says one of his goals for the course is to "teach the language of statistics". However, it is not clear precisely what the language of statistics is. As Coursera's new offerings suggest, statistical reasoning involves multiple facets. In addition to the mathematical side of statistics, which lays out a formal notation for probabilistic models and their operations, there is the ability to effectively describe statistical concepts with words and identifying practical examples where these concepts are at play in the real world, as well as the ability to translate statistical ideas into computer code so that they can be applied and further studied. Thus, there are three distinct yet equally fundamental aspects to statistical thinking. And finding a way to cohesively integrate all three of these perspectives in the classroom has been the greatest challenge educators of statistics have faced, whether in the familiar lecture hall or on the Web. Despite their opportunity to break away from convention, Coursera's classes at this point appear to be following tradition by taking the divide-and-conquer approach to dealing with the multifaceted nature of statistics.
I imagine that many of the 75,000 students who have enrolled in Conway's Statistics One want the whole package wrapped in a radical bow. The astounding enrolment figures are nothing less than an appeal from students for statistical instruction that is bold and path-breaking. So far, there are signs that other online statistical classes have fallen short of the mark. Even with all his appealing zeal, Udacity founder Sebastian Thrun found it difficult to keep students engaged in his course ST101, the first experiment in a no-fee introduction to statistics on the Web. Postings in the class's discussion forum indicated that many students found the work too elementary, giving their comments headings like "do they think I'm dumb??" and "more problems sets please". Other students wanted more programming material and made requests specifically for R ("Statistical Software - Why not R?"). However, with such massive classrooms, even a handful of posts are likely to be a skewed view of a minority of students. Attrition is the ultimate indicator of a free online course's success. Stats on the completion rates for ST101 have not been reported, but by the end of week 2, it was publicly announced that only 13% of students were still actively participating in the class.
Currently, the University of Washington is the only school offering college credit for its Web courses. For now, online universities are, by and large, evolving without the carrot-and-stick of traditional tuition-based education. As a consequence, students who are going to MOOCs to learn are doing so for learning's sake. This is why the percentage of those who stay tuned-in, although not a perfect measure of engagement, is a better measure of a virtual class's success than retention rates at four-walled colleges, where students have few alternatives and more sanctions for dropping out. In the democratized online environment of higher education, where universities are like 24-hour theme parks that have thrown their gates open to let all partake in the main attraction--learning, if institutions are to retain students, they have to, more than ever, hear and be responsive to their needs. The decision to incorporate R into each class in their statistics series is an early sign that, when it comes to the student opinion, Coursera's instructors are listening.