The dangers of statistics [rant]


Illustration: Ellen Luptonhttp://www.wired.com/magazine/2010/04/st_thompson_statistics/
Too true my good friend... too true.

I'm sorry to have a bit of a rant article, I so rarely do, but when it involves math, I get upset.  The fact that this one also hurts my Android fanside only fans the flames.

But, I felt I should react in some manner... obviously the post I'm referring to was just link-bait, as can be seen by the hundreds (and by now, thousands) of replies and comments.
So, what's this article?  Well, here's the original source:


I found it via MobileSyrup, and I clicked it to read that apparently BlackBerry has now overtaken Android for the #2 spot in Canada for 'market share'.  With iOS at nearly 50%, BB at almost 30% and Android at just over 21% (you can see below and try to estimate the numbers yourself)


Which is funny, because not that long ago (March), I posted the comScore numbers of marketshare entirely different:
http://www.androidcoliseum.com/2013/03/android-by-canadian-numbers-stats.html

So in a few months the landscape drifted so dramatically?

No.

Let's take a look at the article that's being shared about:

First off:  it's only reporting the Pageviews by Operating System.  And it's doing so reporting via a handfull of website.

Here's ours for the past week:
Image displaying most popular platforms

So, here's the issue - if I load up a site on my phone and liked something.  I could call it up again later and show it to a friend.  That's 2 pageviews for that one OS.  Also, seeing as some browsers on the different OS's out there allow you to change the UAS to mimic a desktop, so they wouldn't register on this 'smartphone' graph.  Maybe that's just rationalization

And then here's the 2nd point: It was data from just one day.
One day.  That's hardly a good amount to reference.

You can pull it up on the StatCounter site and make your own graph for the types and date ranges you want: http://gs.statcounter.com/#mobile_os-CA-monthly-201206-201306-bar

I'm sure you could play with that thing for a while until it comes up with a satisfactory number.  Which if you look at the source, Michael Blair, is a huge BB advocate... so he was just looking for a win.

Which is another point... why look at just those 2 markets (Canada and the UK)?  They're an American company (Seeking Alpha)... but if you look at the American number, the BB stats are so infinitesimal that it's obviously not good fodder for Michael's point.

Back to Homer's comment.  You could look at any statistic and interpret the way you want.  I mean in fact for my work's website, here's for the iOS visit... clearly you see the fragmentation that exists within iOS




Comments