I did not start off in the field of statistics; my first real job was as a diplomat. And from my undergraduate days I recall a professor who taught a cultural history of Russia. He was one of the world’s top experts. Possessed of a tremendous store of knowledge (a leading author in the field, he went on to become head of the Library of Congress), he also was an enthusiastic lecturer, eager to pass on his knowledge. All of it.
As each lecture period drew to a close, his lecture pace quickened. He had so much to communicate that it could not be held by the bounds of a class period. As the bell rang, he continued, faster and faster, treading on the time of the professors who followed him.
Unfortunately, I remembered none of the content. Like most people, I absorb much more when I actively engage with the material – ask questions, check my knowledge with instant feedback, work on projects, etc. We used to make a division between school and real life, but, more than ever, learning is a life-long project at many levels. And there is a distinction between learning and credentialing.
Data scientists face a varied landscape now for professional development. Some are adept at reading the technical literature. Many simply figure things out on their own. Meetups and professional conferences provide opportunities to pick the brains of others and bounce ideas off them. At the other end of the scale, new Master’s programs in analytics offer the prospect of a comprehensive “reboot” and corresponding diploma.
One of the risks of the do-it-yourself approach is what mathematicians call the “local optimum.” You encounter a problem or a puzzle, you work out a solution, but you have no idea whether there might be better solutions. Or you may fail to see that you are asking the wrong question (what statisticians term a Type III error). Or miss a larger context that should be incorporated into the analysis.
MOOC’s have been a real boon to the self-learner: although promoted as courses, they are a bit more like textbooks on steroids. They lack the real-time involvement of an expert teacher, although some are experimenting with incorporating human interaction via crowdsourcing.
With a well-designed university program, you learn from experts who have that broader context in mind, and who can help you avoid the local optimum problem.
Our own Statistics.com resembles a university program, but also incorporates some of the chaos of the do-it-yourself approach. We try to practice the three I’s: Involvement, Iteration, and Insight.
Courses are more “bite-sized” than in a university program, and every week 4 or 5 courses start. Students are involved in selecting specific courses to meet their needs, and involved within each course (asking questions, doing homework and projects, reviewing and sometimes challenging the feedback).
There is some overlap among the courses. Think how a neural net learns – it’s initial pass through is just a stab at the right answers, then it assesses and adjusts. Humans learn the same way – except for those with extraordinary intelligence and a photographic memory, most of us learn through repetition. Wisdom advances as we take second, third and fourth looks at an issue, from different perspectives – iteration.
From involvement and iteration insight emerges. Even instructors who have been teaching a course for years will have regular experiences of student questions that cause them to look at a problem from a new perspective, and gain insight.