Hi everyone -

  First, a brief apology for those who tried to teleconference in -
something was not working and we missed talking to all of you!  There
were several who were not able to connect.  Our first meeting was
organizational and I think we have a kink or two to work out on the
technology. The good news is that I will summarize what happened here
and we will try again next week to make connections.

  David Banks asked me to convene this first meeting and suggested some
topics which I expanded upon in the agenda below.

  Lisa Denogean agreed to work on our web page and Curt Storlie will
assist her as well.

  We had 9 folks at the table with interests that included Bayesian
methodology and theory, sampling, generalized linear models, regression
diagnostics, computational statistics, likelihood and bootstrapping,
Bayesian model averaging, Bayesian p-values, high dimensional
data, model selection, stochastic structural models, queuing theory,
sequential analysis and change point detection, multiple testing and false
discovery rates, storm tracking, reliability, stochastic ordering, and
time series.  I suggest that we might start off next week with more
introductions from those who were unable to connect with us.

  In agenda item 3 below, we felt there was enough in common about the
applications to stay together for at least a few weeks (or maybe the
whole year) in this large group.

  In item 4 I reviewed some of the data sources mentioned at our BOF
session last Tuesday. Relevant suggestions included
 vital statistics data
 emergency room data
 NYC dept. of health data
 hospital admissions data
 Roy Maxiom's data
 KDD cup data
 Richmond crime data
 Israeli data on terrorists
 Enron data
 Cyclotron data
I asked (and AM asking) each participant to try to think of a relevant
data set or two and check on availability. I personally see data
availability as an important issue in plotting our course. Gauri suggested
some bureau of labor statistics relating to post disaster unemployment
(e.g. New Orleans).

  In item 5 and 6, Greg and I talked briefly about the SCAN statistic.
Last week I found some references at
  www.satscan.org
and passed these out. From those I found a nice paper, Kulldorff:
Communications in Statistics: Theory and Methods, 26:1481-1496, 1997.
Lacking electronic copies, I xeroxed copies for attendees and passed these
out as well.  Gauri volunteered to lead a discussion on this article at
our next meeting. I hope we can all give the paper at least a quick look.
Yajun volunteered to give us an introduction to change point methods. The
idea on these presentations is just to get everyone a bit familiar with
the subjects.

  I also passed out David Banks' recent AmStatNews article that I
happened to come across last week (page 49 of the Sept. issue)  You'll
find that it is a nice summary of much of what we heard last week and
has a nice list of names that we can use to look for reference papers.

  I also reminded the group that there may be a Monday AD group set up as
well at some point.

  I forgot to mention the CHANCE article listed below, but have read it
and it is worth looking at as well.

**************************
 Please be thinking of reference
papers and data sets to bring up at next
Thursday's meeting.
**************************
thanks - Dave Dickey

**********************************************************************************

Meeting 1 Agenda.

1.  Introductions -
      Teleconference participants?
      Your background & association with SAMSI  -  fellow, postdoc,
student fellow, participant, interests e.g. Syndromic surveillance
vs. computer intrusions.

2.  Delegate one of the post-docs to start a website and email alias list
for the group.

3.  Lay out the range of the topic:
     Pass out copies of Banks' AmStatNews article (reviews topics)
anomaly detection in computer traffic, in 911 calls, in keystroke latencies,
in syndromic surveillance, and any other scenarios that we should consider.
Common aspects?  Differences?
  (e.g., time series models, detrending, change point detection, scan
statistics, false alarm rate, ... others? ).

4.  Data sources that may be useful.
      What are relevant data sources - can we get them?
      Assignment for next week - think a bit more about this and whether
you have or can get access to a useful dataset or two.
  (e.g. - I can share some data on stock volume response to the 9-11
attacks)

5.  Discuss methodological issues (e.g., how to improve the scan statistic
to take better account of new false discovery rate ideas; whether wavelet
methods are appropriate for some of this; how to use prior beliefs).

6.  Homework assignments--
       Think about relevant data sources - get people to find articles that
are pertinent to one or more of the applications, and ask two of them to
give a short presentation for the next meeting.  (For example, it would be
good to get everyone up to speed on the scan statistic and/or Bayesian
change point methods and/or wavelets.)

            Hand out Kulldorff article and Spatial Scan reference list.
            Article: CHANCE vol. 17 #1 2004 pg. 19-24
                    Syndromic Surveillance:  Is it Worth the Effort?