medpundit |
||
|
Tuesday, October 18, 2005In their study, Dr Polgreen and his colleagues gave 60 doctors and nurses based in Iowa 100 “flu dollars” each. The participants used these to buy and sell shares coded according to the CDC's colours for a particular week in the future, based on how many cases they thought would occur in the state during that week. For example, if a physician saw three young children with flu symptoms in his office, he might sell any yellow shares he had for the following week and buy red ones. Conversely, if no one he saw seemed to have trouble with influenza, he might buy more yellow or green (sporadic activity) shares for each of the next few weeks. Over the course of the flu season from October 2004 to April 2005, 52 participants logged into the market as traders. They were able to buy and sell up to seven weeks in advance. At the end of the experiment, each flu dollar was converted into a real one and given to the participants in the form of an educational grant. During the course of the season, the participants traded 8,072 shares with a total value of $2,392. And it worked. The flu market was able to predict the colour of a given week with 80% accuracy, two to three weeks in advance. With a lead time of four weeks, it was able to get within one colour either side 90% of the time. These results were better than existing methods, despite the fact that predictions were for the whole state while two-thirds of the traders were concentrated in the eastern part of it, near the university. posted by Sydney on 10/18/2005 12:20:00 PM 0 comments 0 Comments: |
|