|
Abstract:
|
Time and cost are important practical factors in designing phase I trials. Complete and balanced incomplete block crossover designs are able to reduce variance for treatment comparisons and save on sample size. Both designs are well-established, with desirable statistical properties. At times, practical consideration may call for shorter study length and further reduction in study cost. To find common ground between the requirements of empirical sciences and statistical principles, some crossover designs can be adapted without too much sacrifice of their statistical properties (i.e., balancedness). In this talk, adaptations that have been used in practice will be presented. Statistical considerations in power/sample size calculation and data analysis also will be discussed.
|
