Abstract:
|
In a clinical study, the primary endpoint may have different distributions in different cases. Thus, the primary analysis could be based on the finite mixture model. In order to consider these different distributions, this paper will present a method of sample size calculation based on the finite mixture model. We assume that the components in the finite mixture model are normally distributed and the corresponding mixing proportions have a multinomial distribution. According to the Central Limit Theorem, the sample size can be determined using sample mean and variance of this finite mixture model, which are calculated through the conditional expectation and conditional variance. This method is applied to an example for showing the impact of different assumptions on the sample size calculation.
|