Student/Faculty Colloquium-September 22, 1999
Department of Mathematics
Texas A & M - Commerce
Student/Faculty Colloquium
Wednesday, September 22, 1999
3:00-3:50pm, Bin-329
Michael Monticino,
University of North Texas (math dept),
will speak on:
How to stop well on the average
Applications of random probability measure constructions to
optimal stopping problems
Abstract:
Suppose that you are watching the share price of your
favorite stock trying to decide when to sell it so as to maximize
the return on your investment. If you knew the probability
distribution of price fluctuations, then the problem of optimizing
your return could be phrased in terms of a classical optimal
stopping problem and an optimal selling strategy could be
determined. More realistically, you may not be so sure about the
price distribution. You might only have some partial information,
like the average return on the stock. How would you select a
good selling strategy in this setting? One approach would be to
determine just how bad each of the available strategies could
perform, and then select a strategy which performs best in the
worst case. However, while the worse case scenario may be
possible, it could be very unlikely. So instead of comparing the
worst case performance of strategies, it may be more reasonable
to evaluate how well a strategy does on the average.
This talk
will introduce a method of evaluating the average performance of
stopping strategies when partial information is available using
recent work by Hill and Monticino (1998, Annals of Statistics, 26,
No. 4, pp. 1242-1253) on non-parametric random probability
measures.