Determining experimental power for planning purposes is, by its very nature, a risky business. Experiments are planned because we don’t know what the results will be! We hope that the means and variances of control and treated groups will be similar to previous experiments, but that can’t be assumed. There is always the possibility that any imposed treatments will affect individuals differently, changing the variation between treatment groups.
Sometimes the data is not “normally” distributed, for example, frequency distributions do not resemble bell-shaped curves. To analyze non-normal data, some transformation is needed to make it appear normal. QQ Plots may be helpful to see if a given transformation will make the data more “normal” and improve the accuracy of probability estimates using analyses of variance. You can use the workbook QPDOL.exe to check if your data would benefit from some transformation to appear more normally distributed.