USING G*POWER SOFTWARE (Calculating Power) BIOSTATS

PART2

Step-by-Step Guide to Assignment 2.2

Using the same data, Week01_dataset.sav (ATTACHED), you used in the Discussion Question from Week 1 and the related descriptive statistics from the Week 1 Assignment complete the following:

2. Using two different effect sizes, perform three power analyses of the sample size computed in step 1 using G*Power. (Assume a 2-tailed independent sample t test with alpha set at .05)

Step 1. Open your downloaded G*Power software program. Make sure you are in the Central and noncentral distributions tab.

Step 2. Using the down arrows in each tab, select “t test” for Test family, select “Means: Difference between two independent means (two groups)” for Statistical test, and select “Post hoc: Compute achieved power – given α, sample size, and effect size” for Type of power analysis.

Step 3. In the Input parameters, locate the Tail(s) button. Using the down arrow, select Two (you are asked to assume a 2-tailed test).

Step 4. In the Effect size d box, place your cursor over the 0.5 without clicking. A pop-up will appear providing values for small (0.3), medium (0.5), and large (0.8) effect sizes. Type in “0.3” in the Effect size d box.

Step 5. If the alpha (α err prob) is not preset to 0.05, type this in.

Step 6. The minimum sample size needed for each of two groups calculated in problem 2.1 was 38. Type 38 into Sample size group 1 and Sample size group 2. The G*Power settings you should have are shown below.

Step 7. Click Calculate.

G*Power Output

Step 8. Switch to the Protocol of power analysis tab to visualize the Output in the Window.

The Output shows that with 38 cases in each group, there will be only 25% power to detect a significant difference if the effect size is small (0.3).

Step 9. Click the Clear button then click Yes to clear the Output window.

Step 10. Re-run the G*Power calculation using a medium effect size. Change Effect size d to 0.5. Click Calculate.

G*Power Output

With a medium effect size, power increased to 58% with 38 participants in each group.

Step 11. Run the last power calculation using a large effect size. Clear the Output window and change Effect size d to 0.8. Click Calculate.

G*Power Output

With a large effect size, 38 participants in each group will provide 93% power to detect a significant difference.