Hi all,

Thanks to Bahjat for a very nice review of the Chance article. It
generated some nice discussion - thanks to Howard for lending his
expertise to the discussion. Some items:

1. Can we apply some of these methods (CUSUM, EWMA, Shewhart) to
residuals from some sort of model?
2. What are reasonable departures from usual patterns to expect under
the types of interventions we are anticipating.
Bertrand had an interesting framework involving flexible modeling of
effects known to be OK then sequentially adding effects of possible
interventions. He suggested responses like level shift _- ramp /
or bursts |
3. How does multiple testing play into this?
4. (Howard)- Might try including a buffer zone between the background
level estimate and the potential increase. This is to avoid the kind of
slowly changing mean that EWMA etc. accomodate. Going into an outbreak,
the effect might be absorbed into this local mean if it is continuously
updated.
5. Myron suggested regime switching models like perhaps the threshold
autoregressions (Howell Tong has a nice book on these)
6. Jim did a neat followup on the AA data - Lisa is going to post it. I
tried to send it this a.m. but not sure the attachment got through.
7. Francisco had an interesting suggestion about group testing methods
that he mentioned to me after class - we might want to bring that up in
a future discussion. Also Curt suggested we might think of hospitals as
lattice points and see if something useful can come from that view.
8. I plan to talk about the unit root problem as scheduled - Bertrand
suggested a little more on basic time series topics might be worthwhile.
(Let me know if there is interest).
9. Please be thinking about directions we want to take. We are set
until pretty near the end of the semester for our meetings so it's a
good time to take stock of where we are and what we want to accomplish.
I'm new at this and will appreciate any suggestions/help.
10. Howard mentioned what sounds like a very interesting set of
references by David Williams in Statistics in Medicine. He also briefly
described the data we have stored on our site.

Looking forward to Susie's presentation next week - thanks for your
participation.

Dave