Ezekiel: For those interested, there's a book first published in 1954 "How to Lie with Statistics" by Darrell Huff (reprint 1993) that is a first class little book. Highly entertaining & illuminating. It's available at Amazon & I'm sure other places. Well worth the $9 price.

I think I still have a copy of this if anyone wants to borrow it; it was required reading in one of my early biostatistics classes, IIRC. There are indeed lots of things to look out for in the use of statistics, as indeed there are lots of things to look out for in the use of the English (and all other) languages. However, while lots of folks seem eager to beat up on poor Mister Statistics, fewer seem keen on attacking the use of various languages....

...OK, maybe that's not entirely true. A fair amount of attention is paid to the use of language in the construction of logical fallacies. I'm just continuing to feel the need to stick up for statistics. They're really pretty cool.

BTW, if you look at something like that Infoworld article on XP, what is included are figures that barely qualify as statistics. Those figures -- averages/means and such that anyone could measure and calculate with a spreadsheet -- are what is called descriptive statistics. They are less fertile ground for the kind of manipulation that Huff suggests. Essentially, if you can look at all the cells in the spreadsheet, you can do the math yourself. As things get more complicated, and as more complex methods are employed, the opportunity to twist statistics increases.

[edit....]

In my vigorous defense of "proper" statistics, it occurred to me that it might appear that I was dissing Huff's book. That is not the case. I don't remember it that well, but any decent book that points out how statistics (actually more often things like sampling and terminology) can be misused is probably a useful tool for folks looking to critically evaluate the "93.6 percent of respondents think USA Today is swell" types of stat-bytes presented by media outlets, companies, government, and other parties.