Joe Blitztein sent around the following graph:

(The x-axis goes from 2000 to 2012 and the y=axis goes from 0 to 120.)

100 statistics majors (this combines sophomores, juniors, and seniors, but still, that’s a lot more than the 1 or 2 or 3 a year we’re used to seeing). At first I was like, whoa! But then I thought, why *not* 100 or even 200 or 300 statistics majors? Statistics is important in itself, it’s relatively easy as far as quantitative majors go, it’s applicable to lots of other areas. The real question should be not, What’s been happening that’s made statistics so trendy lately? but rather, What took so long for this to happen, and why isn’t statistics more popular? Both places where I studied as an undergraduate, statistics was just a subset of the math department, and maybe the only reason I ended up in statistics is that I took a probability course one semester because, at 5pm, it fit my schedule.

How many jobs or careers are underfilled (or filled with less able people) simply because those who might be well suited for them are not aware of their existence? Or not aware of their nature?

In my neck of the woods, who knows about psychometrics? Who knows about test development and design? Where are those programs? Where might people learn about them and what a good path to prepare for them might be?

Why does it just end of being a tiny number of folks who accidentally fall into them, whether or not they are particularuarly well suited?

This is really heartening to see. I hope this is something that we see in psychometrics (echoing ceolaf’s sentiments) as well as statistics. I am one of those accidental psychometrics students, and while I can understand that this is a small and specialized area of expertise, statistics certainly should be much more popular among the quantitatively-oriented young students.

Prof Gelman, like you I ended up in statistics via a probability course (that my computer science professors recommended), Joe’s in fact! I too wonder why it isn’t more popular, and I try to remember my bias against it and pinpoint where that came from. It is likely a combination of things, but one that comes to mind is my disappointment with the AP statistics curriculum in high schools. I went to a public school lucky enough to offer a few AP courses, but I almost wish I hadn’t taken AP statistics. It took a few years and a few really fantastic courses to erase my perception of statistics as a “pseudoscience”, without rigor or solid reasoning. The emphasis on the ability to “lie” with statistics and manipulate data creates a PR problem.

That you even had the opportunity to take AP Statistics…

That’s a really miserable exponential fit. Among other things, if it were exponential and we focused on some subrange and rescaled the graph we should see something similar, but clearly we would not see this if we attended to the range

2000 to 2006 (or 2007, or maybe even 2008). For this range “constant” would be far better under any sensible metric.

IMO this graph strongly indicates a significant qualitative change in 2007 +/- 12 months. Whether a measurement change, program restructuring, a new program of outreach to prospective stats students, or something else I don’t know – but this graph seems very implausible as a plot of merely organic continuous grown in interest over the entire period. Do you disagree or simply find this an uninteresting question? (You suggest the latter, but it’s a bit hard to believe that.)

Bxg:

1. I think Joe was kidding about the exponential fit.

2. What I’d really like to see is similar graphs for other colleges.

No graphs but this article is informative: http://magazine.amstat.org/blog/2012/08/01/prescornerundergradstats/

This interview with Rebecca Nugent is also relevant starting around the 7:00 mark: http://simplystatistics.tumblr.com/post/33892683645/interview-with-rebecca-nugent-of-carnegie-mellon

bxg: In fall 2006, Joe came to Harvard and started teaching Stat 110, our introductory probability class. He’s a pretty good teacher, to say the least, and the size of the statistics concentration has taken off since then. Extrapolate a mere dozen years under the exponential model, and all of Harvard will be statistics majors. ;)

I second j’s observation. I was lucky enough to be in that 2006 class and told other students not to miss out. I’m sure others did the same, and so the word spread!

Agreed, this might be specific to Harvard. I tried taking the intro probability class at Harvard as a history undergrad in 2004, was miserable and dropped out after a few weeks (along with many of my classmates at the time). In 2008 I tried again as a first-year PhD student in the social sciences and loved it so much I decided to take enough extra stats courses to get a masters degree along the way. I’m sure broader trends like the emerging popularity of “data science” as a field has helped, but I think this graph speaks to the importance of a top-notch intro course to attract undergrads (and even grad students!) who might not have any idea what statistics is really about coming from high school.

I’ll conjecture as to why: Animations, charts, data visualizations, geo-coded stats., prediction contests (e.g. Netflix, kaggle), etc. have really taken off only within the last decade or so.

Those things are “cool” and also participatory and accessable. Although not always done by statisticians the perception is that they are.

p-values, factorial designs, ANOVA: now that’s as boring as boring ever was. Some part of the rise in stat majors arises from this shift in perception, I think.

Another factor may be the rise of Big Data.

Why is factorial designs boring? http://cqpi.engr.wisc.edu/system/files/r076.pdf

I wonder how much of it relates to the MIT-Sloan Sports Analytics Conference, which started in 2006.

I’m not sure the MIT-Sloan Sports Analytics Conference is the reason so much as the rise in popularity in sports statistics in general. For example, Moneyball was published in 2003, and likely took a few years to amass its popularity. A conference is not going to excite high school students, but a popular book might (it did for me!)

I also second the earlier comment about AP statistics. It was an awful experience for me and completely drove me away from statistics. However it was a machine learning class that caused me to be interested again, not a probability class.

I think high school math should include statistics, but AP statistics as it is now is a disaster. I think it is a shame that most high school students in the US get exposed to trigonometry, algebra, and calculus with little to no statistics. Most of my fellow students found the math we studied irrelevant and pointless. Studying statistics in high school (with a well-designed curriculum!) would be excellent.

Kevin & Shira

I agree 100% that there should be more statistics in high school, but a question or two:

1) I flipped through Deseciption of AP Stat in a few minutes, and it did not seem totally unreasonable.

Can you explain what’s wrong with that and what you’;d rather have had?

2) Of course, there is the syllabus and the teacher and both count, so perhaps it was the way it was taught?

John Mashey,

Yes, I think it was the way it was taught. I remember asking questions about where the 1.96 came from in a confidence interval, and my teacher saying something like “you just want to make sure it is wide enough to catch the parameter”. There was a focus on memorizing formulas, without any derivations or intuition. Maybe I was spoiled because my calculus teachers were actually quite good at showing some proofs and intuition. But I think it is possible to give high school students more credit that they can understand something more deeply and not memorize (which I actually tend to find more difficult). We need to make sure that the teachers themselves understand the material enough to explain it well.

Statistics is also well suited to both (a) graduate study and (b) employment after undergraduate. For people interested in medical school, business school, and PhD programs in life science, social science, as well as statistics, it seems like a good option. The biggest factor was likely lack of awareness, and the “big data” buzz might have helped there.

I liked Shira’s theory about AP statistics — the course is indeed insipid — but then it occurred to me that AP Economics is also dreadful and there is no shortage of economics majors at any university.

[...] The number of statistics majors at Harvard is growing exponentially (up 10x since 2006), apparently thanks largely to an outstanding prof taking over the intro stats course. Why, it’s almost as if many students don’t start college with their interests and skills set in stone, and will take to quantitative subjects if they’re taught well! My undergrad college had much the same experience a couple of decades or so ago. Math back then was a very small department, with few majors. But the department hired a number of truly outstanding teachers, and today math at Williams is quite a popular major, and one of the most popular double majors on campus. [...]