|
Activity Number:
|
163
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 9, 2004 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #301966 |
|
Title:
|
Bootstrap Approach for Computing Standard Error of Estimated Coefficients in Proportional Odds Model Applied to Repeated Assessments in Psychiatric Clinical Trials
|
|
Author(s):
|
Qiuhu Shi*+ and Young Zhu and Jianhong Lu
|
|
Companies:
|
New York Medical College and Janssen Pharmaceutical Products, L.P. and Data Solutions, LLC
|
|
Address:
|
School of Public Health, Valhalla, NY, 10595,
|
|
Keywords:
|
bootstrap ; proportional odds model ; clinical global impression ; positive and negative syndrome scale
|
|
Abstract:
|
In this paper, bootstrap method is described for computing standard error of estimated coefficients in building a proportional odds model to explore clinical relevance of Positive and Negative Syndrome Scale (PANSS) in schizophrenia patients. In our model, patient's Clinical Global Impression (CGI) was categorically rated as an outcome variable. Patient's PANSS total score was treated as a continuous independent variable. During a clinical trial, CGI rating and PANSS scale were typically measured more than one time. It was certain that using all measurements would a be benefit in building a prediction model. There were two potential problems. One was that measurements within patient were usually not independent each other, another was that there might be large proportion of missing data. GEE with repeated measure method might not work properly for large proportion of missing data. However, the estimated coefficients in proportional odds model with this type of data were still unbiased, but not for standard error of estimated coefficients. Bootstrap approach on patient level would provide a method to compute standard error of estimated coefficients.
|
