Title: Optimal Scheduling of Stochastically Independent Tests
(joint work with Endre Boros)


Abstract: For many situations of interest (disease, narcotics, steroid
use, nuclear contraband) there are diverse tests that can be applied to
any specific case.  Each test has a cost of application, and a performance
characteristic, compactly expressed by an ROC relation between false alarm
rate and detection rate. In addition to costs of applying the tests, there
may be collateral costs associated with false positives (e.g. biopsy costs
and discomfort, in the case of cancer tests). The problem, under these
conditions, can be formulated as a very large linear programming problem.
With the additional assumption that the randomness in the performance of a
test is due to the hidden variation in the cases themselves (rather than
some randomness in the observational process), the problem can be
dramatically reduced to a dynamic programming problem. The formulation,
solution method, and a discussion of some implications and open problems will 
be presented.


This research is supported in Part by the United States Department of
Homeland Security administered through NSF Grant  0735910) by the United
States Office of Naval Research under Grant Office of Naval Research 
(Grant N00014-05-1-0237), and U.S. Department of Homeland Security through 
the Center for Dynamic Data Analysis for Homeland Security administered 
through ONR grant number N00014-07-1-0150 to Rutgers University.


Bio Note: Paul Kantor is Prof. of Information Science at Rutgers, the State 
University of New Jersey, where he is a member of the LIS department, of DIMACS, 
RUTCOR, and of the CS Graduate Faculty. http://scils.rutgers.edu/~kantor