Figure 3
(A) Effect of sigmoidal dose–response transformation on power for various sample sizes. The simulation power analysis based on transformed dose–response data (simulation power) is slightly less than the power analysis based on MATLAB calculations using the sampsizepwr function (theoretical power). The theoretical power analysis does not take into account a dose–response transformation. Group mean difference = 0.1, sigma = 0.2, EC50 = 0.5, slope = 1, MaxResp = 1, MinResp = 0, alpha = 0.05, nSimulations = 10,000. (B) Effect of slope variability [varSlope; Formula (2)] on sample size for Level 2 simulations at various group mean differences. The difference between the simulation sample size estimates and the theoretical sample size increases at a given group mean difference when greater variability is added to the slope parameter by increasing the varSlope value, with alpha and power held constant. Note that the curves for “Theoretical n,” “varSlope = 0,” and “varSlope = 0.1” are virtually identical and are indicated together. (C) Similar to (B) but for various sigma values.
