Abstract:
|
Mixture model is commonly used to identify subpopulations that may have distribution different from the overall population. Finding the factors that influence the membership of subpopulations and then fitting an appropriate model for the overall population help better understand the data pattern, and therefore make statistical inference more consistent. In this presentation we will discuss developing a mixture model for the measure of radiographic progression in psoriatic arthritis (PsA). With the mixture model, the treatment effect between an experimental arm and the control arm can be explained by both a location parameter and a probability of the membership of patients that progress. Additionally, a regression model assessing these two parameters is developed and an optimization algorithm is applied to estimate their coefficients. The validation and performance of this model will be examined using simulation data. The method will also be illustrated by real clinical trial data from a phase III PsA study.
|