Do I need to crash a high school stats class?

November 12, 2015 11:14 AM

Where do English majors go to learn key principles behind deriving mathematically sound meaning from data? I worked for years as a digital editor. Now I'm a "content strategist," which means I'm responsible for proving my new strategy increases reader engagement via different metrics like UVs, shares, attention time, average story finish rate, etc. I'm very familiar with platforms like Google Analytics and Chartbeat. But I want to learn how to see patterns in them there numbers, and I NEED to make sure my insights are mathematically accurate. How can I most efficiently do that without getting a brain transplant? (Example below.)

For a current client, I'm comparing how Oct Stories A performed compared to October Stories B across 5 different metrics. (The data comes from a proprietary analytics tool.)

All stories were distributed during the same time period on the same platform (Outbrain). It took me forever to figure out that I had to calculate the "percentage difference" instead of "pecentage change between the two groups, and that i have to ensure the ratio of both groups is roughly even. (I'm comparing 9 stories against 11.)

Ratios, percentages: this stuff is not rocket science. But I'm aware I don't know what I don't know. What is the most efficient way to get schooled? A math tutor? Stats 101? Excel training? A lot of data classes focus on the platform but not the meaning behind the numbers.

Thanks!

posted by jessca84 to Education (13 answers total)

Honestly, the stuff you need to know won't be covered in most high school stats classes. However this is exactly the sort of thing I learned in the one university level stats course ("The Practise of Statistics") I took. If you can afford to learn this slowly over a few months, I would recommend auditing a first-year stats evening course at your local university.
posted by 256 at 11:29 AM on November 12, 2015 [1 favorite]

Take a basic stats course, somehow. This stuff has been figured out, and you just need to learn it. It isn't really difficult, mathematically, with all the current software that does the enumeration for you. You mostly need to learn the meaning of all the statistical terms and how they fir together. Your English Majorness will suit you well.
posted by djinn dandy at 11:30 AM on November 12, 2015

Oh, yeah, I also meant to add that you only really need about a grade 9 understanding of math to be able to handle a university-level intro to stats course, so don't worry about getting in over your head.
posted by 256 at 11:33 AM on November 12, 2015

Yeah, this stuff would be covered in a college-level Stats 101 course. If that is not in the cards you might get good enough through Khan or Coursera.
posted by General Malaise at 11:39 AM on November 12, 2015 [1 favorite]

You may want to read The Signal and the Noise by Nate Silver.

In "The Signal and the Noise", the "New York Times" political forecaster Nate Silver, who accurately predicted the results of every single state in the 2012 US election, reveals how we can all develop better foresight in an uncertain world. From the stock market to the poker table, from earthquakes to the economy, he takes us on an enthralling insider's tour of the high-stakes world of forecasting, showing how we can use information in a smarter way amid a noise of data - and make better predictions in our own lives.

And/or consider taking a research methods class.
posted by Little Dawn at 11:40 AM on November 12, 2015 [4 favorites]

When I wanted to learn stats, I actually downloaded lectures from Apple's iTunes U, which is for education stuff. You can essentially sit in an entire semester of intro to stats at a top university. That said, what might be more helpful for you is something more geared toward a layman's understanding of stats, rather than all the principles behind it. A book on using stats for journalists could help, maybe.
posted by AppleTurnover at 11:40 AM on November 12, 2015

Best answer: Duke University's statistical analysis & inference course on Coursera is great. Best in class. (Har har.)
posted by deludingmyself at 11:57 AM on November 12, 2015 [2 favorites]

I think that you'll get more out of a statistical inference for the social sciences type of course than from a more math-oriented Stats 101 type class. Its been my experience that learning the math behind stats from scientists who use it (and think of it in a utilitarian sense) is infinitely easier than learning it from folks who love it for its pure aesthetics/just for the math of it.
posted by Exceptional_Hubris at 12:07 PM on November 12, 2015 [9 favorites]

Best answer: I'd agree on the stats for behavioral sciences route. I teach that, whereas my husband is a math stats guy. Their books spend more time on probability and less on hypothesis testing in practice. A research methods course would be more applied, and particularly one that uses a software program you'd be using would be helpful.
posted by bizzyb at 12:31 PM on November 12, 2015 [1 favorite]

Best answer: Nthing finding a stats course for social science folks. I did two semesters of that in grad school and while they were easily the most grueling courses I took, they were also the most valuable. I can't imagine trying to learn this stuff on my own (I was an English major and art minor), but maybe that's just me? We used Agresti & Findlay's Statistical Methods for Social Sciences for the first course and I thought it was excellent. (Check out the 3rd edition! Much more wallet friendly.) The applied analyses made all the difference.

And for me, the hardest thing was learning how to properly word findings and observations so that they are accurate; writing properly about stats responsibly takes practice. I highly recommend The Chicago Guide to Writing About Numbers.
posted by smirkette at 3:02 PM on November 12, 2015 [6 favorites]

Best answer: Carnegie Mellon has some pretty good free online courses that are really well done. They're not MOOCS and you do them at your own pace--but that also means there's no instructor/built-in community.

The one I've worked through (well, most of it) is Statistical Reasoning (also listed as Probability and Statistics, but both courses appear to be the same).

I think it could be a really good fit for you. The course is introductory, strikes a pretty good balance between theory and application, and I found it clear and informative. You do exercises with real data. If I remember right they have instructions on how to do these in either R (not a picnic if you have no programming background) or Excel or maybe other software packages. One caveat is that while I'm in the qualitative social sciences, I did a few math courses (Linear Algebra, Differential Equations) in college and used to be good at math--so I might have been starting from a different background than you are.

If you really need the basics, you might find it useful to play around with the lectures and exercises at Khan Academy. I gather some pedagogues really disapprove of Khan's style, but for use as a casual refresher I've found it handy.
posted by col_pogo at 3:05 PM on November 12, 2015 [3 favorites]

Eminent statistician John Tukey was an advocate of what he called exploratory data analysis, which means looking at your data in as many different ways as possible until you notice something. Using a statistical software package you can graph data in different ways with a few mouse clicks. Don't let the data intimidate you. Look at representations of it; get familiar with it; what do you see?
posted by Sir Rinse at 5:39 PM on November 12, 2015 [2 favorites]

Response by poster: You guys rock!! Thanka for pointing me in the right direction! "Statistical reasoning" is exactly what I need. I'm going to check out the CMU and Duke classes, and I just bought Smirkette's book recs. :)

(This stuff is actually fun to learn about; who knew?)
posted by jessca84 at 1:25 PM on November 23, 2015