Net Applications numbers are far more accurate than the 3 in the table above.
This strikes me as curious, as I don't know of any popular websites who would agree with that statement, since their own traffic more closely matches that of the wikipedia numbers.
Can you explain why you'd make that claim?
Whether that makes them more accurate? I really don’t know. But I could very well imagine that for those tracking firms their sample is heavily US and Europe biased (where different browsers as in the rest of the world might dominate), so doing some weighing may increase their accuracy. Or not. I really don’t know.
Some light googling does reveal that this is a highly contentious topic with people from all sides screaming at each other about questions of methodology. Do with that info what you will.
On the flip side, Chrome usage dominates among visitors on my development-focused blog at 61.7% and IE trails Firefox (22.2%) at 9.5%.
On the other hand, Net Applications attempts to measure _users_. This is a very different number from visits; it turns out that Chrome users visit more web pages on average than IE users (in addition to a variety of other demographic differences between the two groups), with Firefox users somewhere in the middle. See my example at https://news.ycombinator.com/item?id=7490952 for a simple rundown of how the numbers can be very different depending on what you're measuring, even if we assume perfect measurement.
And I say this as someone with my own large tech-focused site (over 2m visitors a month) which sees about a 40-34-17-3 split for Chrome-Firefox-IE-Opera.
There's been some progress, though, as at least we have fewer clueless people citing W3Schools' stats these days.