An a priori power analysis helps you identify a target sample size. In this blog entry, you’ll learn how to use G*Power to conduct an a priori power analysis for an independent samples t-test.
Setup
First, make sure the following fields are highlighted in G*Power:
Input Options: Tail(s)
In terms of tails, your options in G*Power are one or two tails.
Here’s an example of a hypothesis structure that is one-tailed:
H0: Male income ≤ female income.
HA: Male income > female income.
Here’s an example of a hypothesis structure that is two-tailed:
H0: Male income = female income.
HA: Male income ≠ female income.
Select the number of tails in G*Power depending on which hypothesis structure is present in your analysis. Let’s assume a one-tailed approach for the remainder of this demonstration.
Input Options: Effect Size
Next, consider effect size, which is given as Cohen’s d:
The number you input in this box is the effect size you expect. 0.5 is the default, because it represents what Cohen described as a medium effect size (0.2 is small, and 0.8 is large). Let’s say that, on the basis of past studies, you expect there to be a moderate effect of gender on income; you would therefore leave 0.5 as the default value in this box. However, let’s say you’re carrying out your analysis in a context in which you expect there to be a small effect of gender on income. You would then change the effect size to 0.2.
Input Options: α Error Probability
In the next box, the most common value for α error probability is .05, which is also G*Power’s default.
Input Options: Statistical Power
In the next box, G*Power pre-populates statistical power as 0.95. We want to be able to reject a null hypothesis when it is false, and power represents our likelihood of correctly rejecting a null hypothesis that is false. G*Power’s default power is .95, but .80 is commonly recommended as well. Let’s reset this value to .80:
Allocation Ratio
Allocation ratio refers to the number of people in each of the two groups for an independent samples t test. Leave this value at 1 if your groups are equal. If, for example, one group has 40 and the other has 20, then your allocation ratio could be either 40/20 = 2 or 20/40 = 0.5. It doesn’t matter which of these ratios you enter; it won’t affect the recommended sample size.
Let’s assume that your distribution is 40 people in one group and 40 in the other, so your allocation ratio is 1:
Finally, push calculate.
Recommended Sample Size
Given the inputs we discussed and justified, your recommended sample size is 102:
Again, your recommended sample size will change if you change the inputs. You can follow the advice given above to change your inputs to reflect your own independent samples t test. G*Power will calculate the recommended sample size for you.
BridgeText can help you with all of your statistical analysis needs.