Backwards

As I've admitted before, I'm kind of doing this research thing backwards. I've been developing my own ideas without paying a whole lot of attention to what everybody else has done. My operating premise has been that, if somebody had already cracked this nut, we'd be doing it at work. We have a pretty crack team and so do many of our competitors. There are literally trillions of dollars at stake. And, it's not something that would be kept a secret. Unlike regular insurance companies, which might view such knowledge as a trade secret, reinsurance companies attract clients (regular insurance companies) by showing off how smart we are. People are encouraged to publish.

So, the fact that neither we or our major competitors have figured out a good way to sample our data was enough for me to consider it an open problem.

Imagine my dismay when I came across a paper today that appeared to have addressed it back in 2008. It's one of those moments when you first aren't sure you read that right and read it again. Then you read it slowly while feeling that urge to be sick build in your gut. The fact that it came from of team at University of Illinois was less than comforting; very little that comes out of that engineering school is bogus. Then there's the frantic search to find an online copy of the paper (I'm sure UMSL's library has the journal, but I didn't want to spend an hour going over there this evening). I finally found a copy and was relieved to see that their premise is fundamentally different from mine.

Basically, they set about showing how to construct confidence intervals when your OLAP cube is populated with a sample of the full population. That's a valid thing to think about and you could use it to solve our problem if you were OK with truly random sampling. But, we're not. Random sampling is hugely inefficient. Block sampling is the underlying approach of my paper and that wasn't even mentioned in the U of I work. So, we're all good for now, but it sure did have me in cold sweats for a bit.

I really need to take my literature review seriously this summer.