Abstract:
|
A clinical trial is generally designed to evaluate treatment effect via repeated measures per a short period even when the primary interest is a sustained treatment effect over a long period. In many scenarios such as that the response variable is a patient report outcome, a variable derived via averaging across sampling times and/or items is used in the efficacy analysis. This procedure, however, cannot provide investigators analytical depth and dynamics of the data. Moreover, using averages of non-missing observations per unit time of interest or item to impute missing data may lose non-ignorable information and could introduce biases in the analysis results. Hierarchical models may be used to handle missing data in a different way, help avoid the need of averaging across sampling units in derivation of the variable of interest and maximize the information included in the analysis. Given these potential merits, we explored the hierarchical model and particularly the scenario with an item-response model as one component of the hierarchical model when analyzing the data that are composed of measurements across multiple items and collected repeatedly over a certain period.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.