Power and Sample Size Requirements, Paired Samples

Power depends on expected difference under the null hypothesis (�0, normally set to 0), expected mean difference under the alternative hypothesis (�a, a difference worth detecting), standard deviation of differences (s), the alpha level of the test and whether it is one-side or two-sided (assume alpha = .05 and 2-sided, unless stated otherwise), and the number of paired observations (n). Use web calculator to determine power.

Sample size requirements can be determined in one of two ways. Approach A is to state acceptable margin of error �d and then calculate n = 4s2 / d2 . Approach B entails doing a power calculation in reverse, stating the desired power and then calculating the sample size requirement with the Web calculator.