>Maybe for scientific method or geometry theorem proofs. But not if you're dealing with statistical analysis of trends. There will be exceptions to any pre-conceived theory that you may have when starting to research a question, but the points on the graph will *sometimes* lead to convincing conclusions.

Speaking as a statistician, I don't think you'll find much here. The biggest single problem with surveys, aside from bias, is misinterpreting results. Nancy's point is good. Usually, you want to start out with a strict hypothesis, and attempt to demonstrate it's not likely untrue.

As I noted before in my points, you might be better off just trying to generalize some reasonable theory about the UT/web populations, in a non-scientific manner.