Monday, March 09, 2009

The trouble with statistics #uolstats

R This term I have been mostly teaching statistics. And it's been, well, a bit bimodal really. Half of the students are doing well. The other half aren't, and they're not happy. Which means that I'm not happy either, because I'm sure there has to be a better way than this. It's far from clear that even the students who can jump through the SPSS hoops I set for them retain any useful analytical skills, so I'm pondering how I can do this better.

One possibility is to teach statistics like a pirate: use R

While R has some attractive features, (open source, free, already available on our campus network), on the face of it, abandoning SPSS and possibly Excel in favour of R to improve statistics teaching seems a bit eccentric. For those who don't know, R is a front end for the mathematical programming language S. It operates from a unix-like command line structure, which will scare many students to death. No menus, no glitz. The plan would be to give students an R crib sheet and a statistics decision tree, then turn them loose on the data, assessing understanding by a mixture of numerical and theoretical questions. By stipping out the comfort factor of a familar looking GUI, would students be forced to engage with statistical principles, and ultimately, emerge with more understanding?

To get some perspective on the problem, we're getting together for a little chat this Friday (13th) at 1pm - and you're invited: #uolstats. We'll Twitter the meeting and report back here afterwards.

In the meantime, what software do you use for statistics teaching and why?