Title
|
Data Augmentation Methods
|
Date / Time / Room
|
Sponsor
|
Type
|
08/12/2002
2:00 PM -
3:50 PM
Room: H-Morgan Suite
|
Section on Statistical Computing*
|
Topic Contributed
|
Organizer:
|
David A. van Dyk, Harvard University
|
Chair:
|
Xiao-Li Meng, Harvard University
|
Discussant:
|
|
Floor Discussion
|
3:45 PM
|
Description
Highly-structured application-specific statistical models that aim to account for complexity in both the underlying substantive process and the data collection mechanism (e.g., missing data) are increasingly applied in the physical, biological, social, and engineering sciences. Data augmentation offers a coherent and fruitful strategy for overcoming computational challenges for many such complex models. We explore a number of recent advances in DA methods, including stochastic samplers and deterministic mode finders, and creative implementations to obtain fast convergence.
|
|