From IBM.com, there's a new article for those out there that have been looking for a good starting point on how to analyze the their web data for patterns/trends/etc - Take Web data analysis to the next level with PHP.

The subtopic of this article says it all - Design your data analysis to go beyond simple raw counts. That's exactly what they show you how to do, with things like sampling, using a nominal scale to rank the results, using a Chi Square statistic to help map the stats, and the code to get it all done. This might sound like a lot of stats mumbo-jumbo to most out there, but it can produce some very nice results from your web data.

Yes, this piece is a bit more of the "high level statistics" than most people will want to go, but it might be worth a shot to see if it helps you detect trends in what your userbase is up to...