Above-average learning: Head First Statistics

Yes, you’re right.  Head First Statistics is really a form of teaching, not learning.  As with any book, you could see it as an extended lecture (660 pages, if you count the appendices).  No way to ask anything, easy to slide past questions or problems by turning the page.

Which is why HFS makes such a great example of a tool (depending on your interests) and such a great example of fun (depending on your mindset).

Those two “depending on” clauses are like the uprights for a suspension bridge.  If you’re not interested in learning about statistics and nothing’s pushing you to do so (like your job or your graduate program), then I’m confident (at the 0.975 level) you’re not going to, regardless of the form in which the opportunity to learn appears.  Much the same is true for the other upright: the way you feel about this particular opportunity. 

HFS has given both ends of that bridge some thought. Click the sample page to enlarge; you’ll see what I mean.

♦  ♦  ♦

What Dawn Griffiths has done is sketch a very high-level picture of the learning goals that HFS supports and the types of people who probably respond well to this approach.  Who’da thunk you could do that without the sacred incantation, “At the end of this course the student will be able to…?”

But–how can you be sure of what you’ll learn?

Hmm… the body-of-knowledge approach to learning.  It’s true: in many fields like statistics, there are concepts, principles, terms, equations, and so on that you’re expected to know.

By “know,” I mean you can agree on a description or definition for X with people who aren’t related to you.  Even if their reaction is, “Well, you could put it that way.”

At the same time, despite the sputterosity of purists, zealots, and cranks with time on their hands, most fields don’t have a body of knowledge; they’ve got a herd. Beyond the most basic definitions (like the difference between mean, mode, and media), no one factoid is make-or-break.

Granted, statistics does tend toward the lots-of-specific facts side. So HFS furnishes some tables.

A “Table of Contents (Summary),” which takes up a little more than half a page.  It’s followed by “Table of Contents (the real thing)” with a page for each of the 15 chapters, plus half a page apiece for the intro appendices).  Check the O’Reilly Books preview page for HFS yourself; use the next / previous buttons at the top of the book page to browse.

If I were smart, I’d end by suggesting you also look at a sample chapter (probability, PDF) or explore HFS on your own via Google Books.  The good-humored approach, the absence of dense text–those are obvious at first glance.

Beneath that, though, there’s a lot of cognitive infrastructure, the sort of thing that shifts from “fun” to “learning.”  Chapter 3, “Power Ranges,” is a good example.  It’s got 44 pages dealing with range and variation (the previous chapter dealt with mean, median, and mode).  This is what’s lurking as you turn the title page:

  • The coach of the neighborhood basketball team needs one player.  He’s got three candidates.  All three have the same shooting average.  So, which one should he pick?
  • Here are their individual stats (points per game and frequency).  What else does the coach need to know?
  • Explanation: what “range” means (also, lower bound and upper bound).
  • You try it: figure the mean, lower bound, upper bound, and rang for these two players.  Then, draw a histogram (as you learned in chapter 2) for each.
  • Feedback for that exercise, and a troubling question about outliers.
  • Explanation: why outliers are problematic.  Can you think of how to reduce their impact?
  • Explanation: why ranges are quick-and-dirty solutions.
  • Sneaky intro (“one way is to measure only part of the range”)
    accompanied by this:

That’s the first 8 pages.  Not only did you have a couple of get-out-your-pencil problems, but also questions to provoke thinking, questions to highlight potential confusion, and even, as in the above example, questions that are intelligent stand-ins for ones a learner might have.

As I said in an earlier post, fun in training (or in support of learning) shouldn’t be an afterthought.  It shouldn’t be force-injected, either, like the fake smoke-flavored streaks applied to frozen burgers to make you think they were grilled.

Dawn Griffiths shows that part of the engagement comes from a general approach (irreverence, retro photos, quirks of layout); another part comes from sample problems that offer real statistical challenges placed in…let’s say surreal settings.  (In chapter 7, you use Poisson distributions to figure out how often a movie theater’s popcorn machine is going to break downnext week.)

Maybe we can get Griffiths and the folks from Head First to have a long lunch with von Merriënboer and Kirschner.