The simulations show that the parameter <i>N_H</i> raised to the power of 0.37 is the one that empirically allows a better fit of the data to the theoretical curve of statistical power even in scenarios where the number of cases differs from the number of controls.

<p>See legend of <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0073567#pone-0073567-g003" target="_blank"><b>Figure 3</b></a> for more information on the simulation.