Since we often count up server usage for Web sites, we are always curious about people's counts. Recently, we saw one on programming languages that was both inspiring and that inspired some thinking about how stats are interpreted -- check it out at http://www.tiobe.com/tpci.htm...
For example, your gut might tell you that PHP might be more popular than this site would have you believe, especially if you count sites with particular file extensions or server responses (remember our friends at http://news.netcraft.com/?). Now, further consider from the survey at http://www.tiobe.com/tpci.htm that some of these languages can't really be used for building desktop apps or rarely found in anything beyond Web applications and that might give you a much different read on the position of say JavaScript in such a survey...
We bring this up for the simple reason that this data has the same old problems with stats: sample size, sample makeup and data presentation.
When we see server stats compared to ours, folks are not always willing to take into consideration that parked domains all running one OS/Server combination seem to be counted the same as a farm of 100s of servers running a certain Redmond, WA firm's software (think Barbarian, I mean, Bill at the Gates).
If this post piques your curiosity, an enlightening and fun read on stats (complete with pictures) is the Darrell Huff classic "How to Lie with Statistics":
http://www.amazon.com/gp/product/0393310728/104-4742397-5997557?v=glance&n=283155
You'll rarely pay attention to any poll numbers after you read this book!
Cheers,
Port80 M.F.