|
Abstract:
|
Purpose: The study aims to highlight valuable and informative simulation data to improve power analysis and sample size justification. Background: Typical power analysis and sample size justification often provide summary statistics to argue the proposed sample size with sufficient power to detect the hypothesized effect size through analytic formula. The concrete summary could be more informative if variation of parameters of interest by simulation is incorporated to better assess uncertainty of effect size. Method: A Monte Carol simulation based statistical approach is developed to evaluate variation of primary parameters. The approach delivers valuable information regarding distribution of relevant parameters to evaluate the effect size. Result: A two-arm randomized design is used for illustration with survival time as endpoint and hazard ratio (HR) for effect size. Under the alternative hypothesis of more effective in the treatment (i.e., HR< 1), simulation does not only calculate the power, but also demonstrate variation in parameter estimates of interest even with unexpected results. Conclusion: Incorporation of simulation results could strengthen the statistical plan.
|
