|
Activity Number:
|
314
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 11, 2004 : 9:00 AM to 10:50 AM
|
|
Sponsor:
|
Social Statistics Section
|
| Abstract - #302090 |
|
Title:
|
Analysis of Messy Longitudinal Data from a Study of Pediatric Patients
|
|
Author(s):
|
Shesh N. Rai*+ and Shelly Lensing and Sean Phipps and James Boyett
|
|
Companies:
|
St. Jude Children's Research Hospital and St. Jude Children's Research Hospital and St. Jude Children's Research Hospital and St. Jude Children's Research Hospital
|
|
Address:
|
Dept. Of Biostatistics, Memphis, TN, 38105,
|
|
Keywords:
|
longitudinal study ; quality-of-life study ; model-based inference
|
|
Abstract:
|
"Messy" data are more common than "neat" data in real-world clinical studies. The quality of messy data is always questioned; therefore, inferences drawn from such data need special attention. We describe a messy data problem that confronted us and the way that we dealt with it. The objective of the study that generated the messy data was to relate baseline psychometric status to longitudinal quality-of-life indicators for pediatric patients with cancer. The data were very messy because of the imprecise manner in which they were collected and their incompleteness from a variety of sources. Furthermore, analyses of these data were complicated for several reasons. The longitudinal nature of the data required modeling of the correlation between multiple measurements taken on the same subject. Missing responses necessitated justification of assumptions. Different analytic approaches were used because continuous and ordinal dependent variables were obtained. Simplification and grouping strategies were used to deal with the large number of independent variables. We discuss the model-based inference procedures used to address this problem and describe the application.
|
