|
SPEAKER: |
Dr. Peter Song, York University |
|
TITLE: |
"Marginal Models for Longitudinal Compositional Data" |
|
DAY: |
Wednesday, October 22, 1997 |
|
TIME: |
3:30 p.m. [Coffee in BSB-202 at 3:00 p.m.] |
|
PLACE: |
BSB-108 |
Marginal models with exponential dispersion model margins proposed by Liang and Zeger (1986) have become a useful tool for analyzing longitudinal continuous or discrete data. They provide the basic needs for practitioners to understand a population-averaged cause-and-effect mechanism between the response and covariates. In this talk I shall present marginal models for longitudinal compositional or proportional data whose marginal distribution is assumed to be the simplex distribution of Barndorff-Nielsen and Jorgensen (1991), which is not included in the family of exponential dispersion models. Liang and Zeger's generalized estimating equation approach is extended to deal with parameter estimation for our models. An analysis of gas volume decay in eye surgeries is presented to illustrate the proposed method.
Dr. Peter Song is an Assistant Professor in the Department of Mathematics and Statistics at York University. He obtained his Ph.D. from the Department of Statistics at University of British Columbia in 1996. His research interests include generalized linear models for multivariate data, state-space models for multivariate time series of non-Gaussian observations, stationary time series models with non-normal margins, and statistical computing.
The following articles have been placed on reserve at Thode Library (STATS 770: Statistics Seminar). They have been provided by Dr. Peter Song to be used as background for his talk on Wednesday, October 22.
Unfortunately neither of the books is in the library. If any one receiving this message has one or both and wishes to lend them for a few days, just let me know. The students interested may then borrow them.
[1] Diggle, Liang and Zeger (1994). Analysis of Longitudinal Data, Oxford University Press.
[2] Jorgensen, B.(1997). The Theory of Dispersion Models, Chapman and Hall.