|
Activity Number:
|
380
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #303427 |
|
Title:
|
Sample Size Calculation for a Mixture of Discrete and Continuous Endpoints
|
|
Author(s):
|
Yolanda Munoz Maldonado*+ and Sarah M. Baraniuk and Lemuel A. Moye
|
|
Companies:
|
Michigan Technological University and The University of Texas Health Science Center at Houston and The University of Texas Health Science Center at Houston
|
|
Address:
|
1400 Townsend Drive, Houghton, MI, 49931,
|
|
Keywords:
|
poisson distribution ; binomial distribution ; multivariate normal ; asymptotic distribution
|
|
Abstract:
|
Sample size calculation is a key component in designing experiments. Typical practice for calculating sample sizes in design of experiments involves testing differences between a treatment and a control group on one primary variable, either discrete or continuous. However, the sample size calculation becomes more complicated when trying to test hypotheses for a mixture of discrete and continuous variables that are correlated. In this talk, a Poisson-Binomial distribution is used in combination with the multivariate normal distribution to construct a composite endpoint. We provide an example using data from the CARE trial. A Montecarlo simulation is performed to investigate the critical values and power of the procedure and the sensitivity to some of the assumptions involved.
|
