The objective of this thesis is to acknowledge some of the different randomized trial designs and analytical approaches, as well as specific issues related to these. The thesis can be divided into two parts. The first part focuses on the validity of using randomization tests and tests based on meta-analysis techniques in the context of cluster randomized trials with binary outcomes. Reasons for making statistical inferences based on a randomization model rather than on a population model are discussed. Also, the theory of randomization tests and tests based on meta-analysis techniques is reviewed. The validity of using such tests is evaluated for various study designs through simulation of size. Finally, application of the theory is made to the CHAT program. In the second part of the thesis, the appropriateness of using multiple personal digital assistants (PDA) per centre, each with pre-loaded blocked randomization schemes, to allocate participants in multi-centre randomized trials is evaluated. The probabilities of allocation imbalance and their potential effects on power are evaluated by simulation.
展开▼
