We have a few tests of the effect of major, globally monitored sporting
events, notably World Cup Soccer. They have generally not produced
persuasive effects on the EGG network.
Jay King wrote on Jan 26, "I was wondering if the project plans on
checking the GCP for any events
occuring during the time of the suprbowl. That is certainly an event
that is monitored by people all over the world."
I responded, "If you give me beginning and ending times, I can at
least do an exploratory look. It might be worth the
investment of a formal prediction, but is several test
cases, we have not seen significant effects for sporting
events (e.g., World Cup Soccer, which is even more popular
than the Superbowl.)"
He did write back and suggested the period from 6:00pm and 11:00pm that
night (The game was in San Diego, but TV guide had east coast show time from
18:00, so that's 23:00 on the 26th to 04:00 on the 27th GMT). This
graph shows the result, for 50 Eggs. There is a strong trend at the
beginning, but it falls off to a level random walk during the second half.
The terminal Chisquare
is 18152 on 18000 df, which is equivalent to a probablity of 0.212.
Bryan Williams of the University of New Mexico also expressed interest
in the Superbowl. He has done some analyses and explorations of GCP
data and said he would take a look at the Superbowl data using 15-minute
blocking, but with an otherwise similar prediction to mine.
The following figure shows his result. He reports that the negative
going trend has a Chisquare of 896.25 on 944 df and p = 0.864.
This is contrary to expectation, but is not statistically significant,
despite the relatively steady trend.
|