|
Activity Number:
|
75
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 6, 2006 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #307042 |
|
Title:
|
Recent Development in Exact Inference for Parallel Group Design with Repeated Binary Measurements
|
|
Author(s):
|
Dar-Shong Hwang*+ and James Lee+
|
|
Companies:
|
B.R.S.I. and Sankyo Pharma Development
|
|
Address:
|
11 Trinity Place, East Hanover, NJ, 07936, 399 Thornall Street, Edison, NJ, 08837,
|
|
Keywords:
|
exact conditional distribution ; correlated binary variables ; generalized constant odds ratio ; generalized treatment effect ; generalized interaction effect ; parallel group design
|
|
Abstract:
|
Parallel groups with repeated binary measurements occur often in clinical trials. Hwang & Lee(2005) investigate an exact analysis of treatment*time interaction and treatment effect for 2 treatment groups with 2 or 3 repeated binary measurements. This paper further indicates the difference in the definition of interaction, treatment and time effects between repeated or independent binary measurements. The treatment effect assuming no treatment*time interaction is also clarified. Analogous to Zelen's test of interaction, and Birch-Cox-Gart's test of treatment effect assuming no interaction for independent binary data, parameters representing generalized treatment*time interaction, treatment or time effects are developed assuming no generalized interaction or constant odds ratio for correlated data. Existence and optimality of exact inference are demonstrated and actual distribution derived.
|
