User: Password:
Subscribe / Log in / New account

Painfully obvious truths

Painfully obvious truths

Posted Feb 26, 2013 20:56 UTC (Tue) by mathstuf (subscriber, #69389)
In reply to: Painfully obvious truths by man_ls
Parent article: Mozilla announces 18 carriers supporting Firefox OS

Thanks for the links; they're some interesting reading. I'll have to send an email to Ars asking why they use NetMarketshare (and maybe do a piece comparing them with StatCounter). The weighting based on the CIA factbook seems, to me, to indicate that IE is used more heavily in less connected countries (since they get pulled up). And yes, it is weird to weight your data, but, IMO, no less weird than just ignoring missing data (which would seem to be what SC does). At least they both are

> If you have any other enlightening views about why the bias, please let us know.

I suppose if and when Ars does a NM/SC comparison article would help with this.

> Finally, given that I was talking about dwindling Firefox market share, which is confirmed by all sets of data, this is a very unnecessary discussion, entertaining as it may be.

Well, seeing as my comment spawned this subthread, I agree that Firefox isn't looking strong in any of these datasets. I was more surprised at the difference between the stats I had seen before. Plus, it's not like I have much of a horse in the race as a developer of Uzbl (which will forever be relegated to a fraction of "Other").

(Log in to post comments)

Painfully obvious truths

Posted Feb 26, 2013 21:03 UTC (Tue) by dlang (subscriber, #313) [Link]

> ...indicate that IE is used more heavily in less connected countries...

This makes sense. IE is already on the machine, Firefox (or any other browser) will need bandwidth (i.e. time) to download.

If you don't really need it, and surfing the web is slow and painful in any case, why bother?

Painfully obvious truths

Posted Feb 26, 2013 21:03 UTC (Tue) by raven667 (subscriber, #5198) [Link]

> I suppose if and when Ars does a NM/SC comparison article would help with this.

They usually compare the third party aggregate stats with their own site stats which are often wildly different, much less IE, way more Firefox.

Painfully obvious truths

Posted Feb 26, 2013 23:15 UTC (Tue) by Lennie (guest, #49641) [Link]

OK, I'm replying at the bottom of this thread in the hope that most of you subscribed to notifications all get this reply.

Most of the time you see 2 organisations quoted in news articles for statistics. and

hitslink always has a much higher IE-share than statcounter, there are several reasons for that that I know of:

1. hitslink seems to be on more business oriented, IE is bigger in business than at home (just look at statcounter and compare the numbers for weekday and weekend).

2. IE is still bigger in China than in western countries. The country with the largest share of Windows XP is also China I believe. Probably Google isn't such a household name in China as their searchengine does not have a large share. A large part of Chrome users are Chrome users because of Google advertising and them being a household name. The reason China is significant for the statistics is because hitslink mangles their statistics based by adding a weight based on the number of people in a country. And China is a really large country.

3. statcounter measures by pageviews and hitslink counts by 'visit' (per session of pageviews). IE users are the 'dumber' users, users that don't know alternatives exist or how to install new software. My guess is these are users which visit a whole lot less pages per website. Chrome and Firefox users click much faster thus visit more pages.

Counting the missing

Posted Feb 26, 2013 22:44 UTC (Tue) by man_ls (guest, #15091) [Link]

Well, that explains it. I would argue that counting the browsers of people who don't browse is useless and even misleading. But ignoring missing data which you don't have and you don't really know how much it is worth, is wise. It would be like trying to guess what browser would have all those people that do not have computers, if they bought computers -- and adding them to the stats.

Counting the missing

Posted Feb 26, 2013 23:07 UTC (Tue) by mathstuf (subscriber, #69389) [Link]

I suppose I should have been more clear: it's weird in the sense that there are no error bars on anything. One tries to compensate for the missing data errors while the other says "what we have is good enough". Neither approach is fundamentally good or bad under difference circumstances. SC has a much larger base versus NM based on site counts, so NM has to make up the difference somehow to make as strong an argument.

Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds