Wednesday, May 18, 2005

So what does 1/1000 mean?

Today on ATC, Richard Harris did story on an hourly earthquake forecast system in California. He mentioned that once the system predicted a 1/1000 chance of an earthquake before a major earthquake occurred, and thus clearly the system was "wrong."

I wonder what he would have considered "right" in that case. Would it be right if it was 300/1000? Or 750/1000? What if it gave a probability of 900/1000 or even 999/1000 and the earthquake did not occur? Would the system be "wrong?"

Later in the report he mentioned that this system has been more accurate in its predictions than weather forecasting systems because so few earthquakes have occurred. I wonder how he determines when a weather forecasting system is "right?" Apparently the earthquake system is "right" when it predicts 1/1000 and an earthquake does not happen, but when a weather forecasting system predicts a 40/100 chance of rain and it rains, then the weather forecasting system is "wrong."

Fascinating stuff.

No comments: