Friday, September 26, 2003
300 Yost Hall
Talk: 4:00 -- 5:00 p.m.
Refreshments: 3:30 -- 4:00 p.m. in 300
Yost
The detection and estimation of multiple covariance matrix shifts for a
sequence of m dimensional ( m > 1 ) Gaussian random vectors by using the
Schwarz information criterion (SIC ) will be discussed. We will estimate
the number of shifts as well as their locations.
The unbiased SIC's are also obtained. Then the asymptotic null
distribution of the test statistic is derived. The results are also
extended to the general non Gaussian case. Then the results are applied
to the weekly prices of Exxon and General Dynamics stocks from 1990 to
1991, and the changes are successfully detected.