An interesting observation

[Mathiasdm]

Those stats don't even look right: iPhone has 0.4% market share? extrapolated that means that there are 1.1 billion internet users (http://www.internetworldstats.com/stats.htm) and Apple have sold about 700,000 iphones (that's the biggest estimate I could find) yet your link says there are 4.4 million iphones, and that's assuming every iphone sold has visited those sites (which I don't buy either).

You might have a point there (though you did make one mistake with the iPhone stats: it's 0.04%, not 0.4%), which would amount to 440 000 iPhones. The question is: how many % access the internet using their iPhones?

So I consider those stats to be based on very small numbers, even if your getting logs from 40,000 urls, if only 1 visitor visits half of them (20,000 requests) and say most of those urls deal with windows news or Mac blogs, something linux users wouldn't be interested in. it would cause some very weird results, and mostly I bet to show how low the linux market share is. I think those stats have been designed to show linux in a bad light.

It seems quite normal that some websites in that list will have mostly Windows or Mac users (due to being aimed towards them), but there might be Linux websites in that list as well. You just can't know.
'have been designed to show Linux in a bad light' --> let's keep the conspiracy theories away unless you have proof

For normal websites, ecommerce websites, email websites (all the places I've worked behind the scenes) they all show between 2.5% and 4% market share to linux. and I would believe an ecommerce sites statistics (like bbcshop.com) before some random urls.

I started looking for more sources, so:
-XitiMonitor: 0.8% Linux market share
-OneStat: 0.36% Linux market share
-W3Counter: 1.32%

Each and every one of these websites tracks a whole lot of sites, so gives a more reliable estimate of market share. That's why I'm guessing Linux market share is not at 3%, but around 1%.

The annoying thing is that it's very hard to track Linux market share

[Zzl1xndd]

Well Mac market share is pretty easy given we just need to look at the numbers sold and it Looks to be about 5% total but I have seen some reports that they now control 15% of the notebook market perhaps a sign of things to come. Also Linux market share is normally thought to be in the 3-6% range but even that is a guess at best as site stats don't mean much as I have it set up to trick sites into thinking Im using IE on Windows XP even though Im not.

[mangar]

Does those stats take into account that many (I don't know if most, but most certainly Google's) search index crawlers uses Linux? In this case, the actual market share may actually be lower that 0.36%.

[a12ctic]

The reason you never see anybody wearing windows shirts is because almost nobody uses windows out of choice, they either dont know better or are tied to an application.

[Mathiasdm]

Does those stats take into account that many (I don't know if most, but most certainly Google's) search index crawlers uses Linux? In this case, the actual market share may actually be lower that 0.36%.

That's actually a good point.
But, while regular web browsers have their OS listed in the browser string, this might not be the case with web crawlers. It's apparently possible to check if a visitor is a robot, by seeing if they access the robots.txt file

[Depressed Man]

I’ll let you secret… if you have an o/s logo on your shirt everyone will still think you’re a loser no matter what one it is.

I doubt most people would know what the Ubuntu logo is, much less the Penguin that represents Linux. While with OSX (actually what is their logo? The Apple symbol?) and Windows is easier to recognize since they have a marketing presence.