|
Activity Number:
|
111
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #305708 |
|
Title:
|
Modeling W-Shaped Data
|
|
Author(s):
|
Robert J. Gallop*+ and Randall H. Rieger and Scott McClintock
|
|
Companies:
|
West Chester University and West Chester University and West Chester University
|
|
Address:
|
24 Rampart East, Media, PA, 19063,
|
|
Keywords:
|
Zeroes ; skewed data ; zero-inflated models ; mixtures ; repeated measures
|
|
Abstract:
|
Regression analyses will only provide correct inferences when certain assumptions about the data are met. When the assumptions are not met, a common remedy is to seek a transformation that leads to normally distributed residuals. It is quite common to find a large proportion of the data stacked at one response such as 0. No transformation can spread out a stack of zeroes. ZIP/ZINB regression models have been used to model such data. When the data is stacked at two extremes with a near normal distribution in between, resulting in a bimodal W-shape distribution, a new modeling structure must be used. We propose a mixture model in which the complete distribution of the outcome is approximated by mixing three component distributions: the two extreme responses and the data in between. We compare our results to the standard modeling approaches and discuss extensions for longitudinal data.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2009 program |
