Many software packages will calculate power for you, especially with the normal approximation to the
Calculating Power (Normal)
function uses the following inputs
= the desired level of significance (
= the sample size (number of trials)
= the process probability (
) under the null hypothesis
= “less”, “greater”, or “two.sided”
= the alternative probability of success
> iscamnormpower(LOS=.05, n=20, prob1=.25, alternative =
should reveal both distributions and report the rejection region to achieve the level of significance, the
observed level of significance, and the power.
Stat > Power and Sample Size > 1 Proportion
Specify the sample size(s), specify alternative values in the first two boxes.
Specify the hypothesized probability in the last box.
specify the direction of the alternative and the level of significance.
Keep the Distribution pull-down menu set to Normal and specify the values for the mean and
tab. Define the Shaded Area by
, specify the tail direction,
and enter the observation value of interest. Press
Power Simulation Applet
There is a
Normal Approximation check
In fact, the most common application is to specify the desired power and solve for the necessary sample
size before conducting the study to determine how many observational units you should recruit.
(e) If your technology allows, see how many sessions would be needed in the ESP study to have at least
an 80% chance of rejecting the null hypothesis if the actual probability of success is
(f) How will your answer change if the actual probability of success is