"Lies, Damned Lies, and Statistics"

In class this week we looked at the Gapminder website which allows people to manipulate data to look at "relationships" between different variables.  I enjoyed playing with the data and the site has a big "ooh" and "ahh" factor.  The main issue is having enough of an understanding of how statistics works to be able to critically appraise the data you are seeing.  I like Mark Twain's quote  "There are three kinds of lies: Lies, damned lies, and statistics."  In research (and, lets not forget, politics), it is pretty well known that you can find a relationship between any two sets of data if you look hard enough.  This is why, in general, researchers start out with a specific hypothesis that they are seeking to prove or disprove based on a (hopefully) carefully structured study design.  Starting out with a question should preclude a researcher from randomly searching through the data they have collected to find a relationship.  A good design should try and control for potential confounding factors or at the very least acknowledge their potential existence.  Additionally, the availability of data and the quality of that data from any one country or source needs to be considered as well.  Certainly, some of the relationships that can be shown and discussed would be perfectly valid. Some, less so.  It is also true that we are all bombarded with data and statistics every day so the ability to evaluate this type of information is critical.  With all of this in mind, I am trying to figure out ways that this type of site can be best used in a classroom.