How to Compute Power Analyses for Sample Size
In statistics, it isn't always possible to survey an entire population, so a sample from the population is used instead. A power analysis will tell you what size sample is necessary to get an accurate result that's reflective of the entire population. In other words, power analyses will help you avoid errors in testing. The sample size should reflect a power of above .8, which means you have an 80 percent or higher chance of detecting statistically significant differences in your data. Conducting power analyses involves a series of complex calculations. Researchers commonly use software programs such as G-Power (a free software program from the University of Dusseldorf) to perform the calculations.
Instructions
-
-
1
Start G-Power (see Resources). Select the type of power analyses. The type of power analyses chosen will depend on if you are calculating before or after you conducted the experiment.
-
2
Select the type of test. The type of test is the best type of test for your experiment (for example, independent t-test, ANOVA or regression are all techniques you can use to find the correct sample size).
-
-
3
Decide what Type 1 error rate you are comfortable with. A type 1 error rate is the amount of error you are willing to accept. For example, you may decide on a 5 percent error rate.
-
4
Decide what your expected effect size is. An expected effect size is basically what difference you expect to find in your data. Effect size is generally found by subtracting the mean of the control group from the mean of the treatment group, then dividing the result by one group's standard deviation.
-
5
Enter a sample size, and G-power will generate a number between zero and 1. Your goal is to have a number above .8; if your number is lower, adjust your inputs and try again.
-
1
