SPEAKER: |
|
TITLE: |
"Inference procedures on inverse Gaussian scale parameter" |
DAY: |
Wednesday, November 8, 2000 |
TIME: |
3:30 p.m. [Coffee & cookies in BSB-202 at 3:00 p.m.] |
PLACE: |
BSB-108 |
The conventional definition of the inverse Gaussian distribution involves two parameters, the expectation mu and lambda which is referred to as a scale parameter. We consider methods for inference problems involving the parameter lambda. It is observed that likelihood based methods for such problems are non-robust in the sense of Box (1953); that is, the associated Type I error control and coverage probabilities are not even asymptotically achieved. In this paper, we consider construction of robust inference methods for the problems with an emphasis on the jackknife technique. The problems considered include estimation of lambda and testing homogeneity of lambda parameters from several populations, both in order-restricted as well as unrestricted settings. Robustness of the procedures is examined for variety of alternatives including lognormal, gamma, Weibull and the recently introduced contaminated inverse Gaussian populations.
Dr Rajeshwari Natarajan received her Ph.D. in 1999 from the University of Rochester, New York, under the supervision of Prof. G.S. Mudholkar. Her thesis made key contributions to the study of Inverse Gaussian distributions and related inferencial issues. She is currently working in the Department of Statistical Science at Southern Methodist University, Dallas, Texas.
The references below, suggested by the speaker as useful background for this talk, have been placed on reserve at Thode Library (STATS 770: Statistics Seminar).