Abstract:
|
Bayesian designs of clinical trials offer flexible solution to accommodate the uncertainty of parameter estimation, the possibility of a reduced sample size and increased statistical power of a hypothesis test. In single-arm clinical trials with a binary response, with application of commensurate priors, power priors and hybrid priors to adaptively incorporate historical data, we investigate choices of the associated hyperparameters and weights. With different sample sizes and binomial probabilities in historical and current trials, posterior mean square errors and posterior coverage probabilities of binomial parameters are compared to reveal appropriate scenarios of using the hybrid priors.
|