## ABSTRACT

This chapter discusses the clinical trials should be designed to minimize potential bias and to balance efficiency and generalizability. On the other hand, some authors prefer to use the term calculation for sample size to distinguish from the estimation of a model parameter in the context of making statistical inferences; these estimated values are usually assumed as fixed in the calculation, and seldom considered for their uncertainty. A randomized trial was conducted to assess the effectiveness of Relenza, a new treatment for influenza, compared with placebo. The primary variable of interest was the number of days to alleviation of symptoms. However, the same formula applies to the more realistic case of random censoring due to staggered entry of patients with the same study duration for each patient. The hazard rate under the exponential model is constant over time. Sometimes, the hazard rate may be constant, but increasing or decreasing over time.