G*Power Sample Size Calculations: 5 Min Demo

Summary

This video provides a step-by-step demonstration of how to perform an a priori power calculation using G*Power, a tool for determining the sample size needed for reliable statistical analysis. The video covers downloading and installing G*Power, selecting the appropriate statistical test (independent samples t-test), setting parameters like effect size, alpha level, and power, and calculating the required sample size. The presenter also explores different scenarios, such as varying effect sizes and study designs, to showcase how the sample size changes. The video concludes with tips on adjusting for one-tailed tests and paired samples t-tests.

Takeaways

• 😀 G*Power is a software tool used for statistical power analysis and can be downloaded from gpower.hhu.de.
• 😀 The first step in using G*Power is choosing the correct statistical test, such as an independent samples t-test for comparing two independent groups.
• 😀 The effect size is a critical parameter to select in power analysis. Commonly used effect sizes are small, medium, or large, with 0.5 being the medium effect size.
• 😀 Alpha level is set at 0.05, which is the standard threshold for significance testing in most studies.
• 😀 Power, commonly set to 80%, represents the probability of detecting an effect if it exists.
• 😀 The calculation for sample size estimates how many participants are needed to detect a certain effect size reliably.
• 😀 For a medium effect size and 80% power, 128 participants (64 per group) are required for a two-tailed independent samples t-test.
• 😀 If a study only aims to detect a large effect size, the required sample size drops significantly, requiring just 52 participants in total.
• 😀 A one-tailed test (if justified and pre-registered) can further reduce the required sample size, bringing it down to 42 participants in total.
• 😀 If a paired samples t-test is used instead of independent samples, the required sample size is significantly reduced to just 12 participants in total.
• 😀 The tutorial emphasizes the importance of conducting an a priori power analysis before conducting research to ensure the study is adequately powered.

Q & A

What is the purpose of conducting an a priori power calculation?

An a priori power calculation helps determine the necessary sample size before starting a study to ensure that the study has enough power to detect a statistically significant effect, if one exists.

What software is being used for the power calculation in the video?

The software used is G*Power, which is a free tool for conducting power analysis and sample size calculations.

How can one download G*Power?

To download G*Power, go to the website gpower.hu.de, scroll down to the download section, and choose the version suitable for your operating system (Windows or Mac).

What statistical test does the video demonstrate for power calculation?

The video demonstrates an a priori power calculation using the independent samples t-test, which is used to compare the means of two independent groups.

What are the default effect size conventions mentioned in the video?

The default effect size conventions mentioned are small, medium, and large. These can be used as a guideline or adjusted based on the specific context of the study.

What is the alpha level typically set to in the power calculation?

The alpha level, which determines the threshold for statistical significance, is typically set to 0.05 in the power calculation.

What power level is commonly used in research studies?

The power level is commonly set to 80%, which means there is an 80% chance of detecting a true effect if it exists.

How does changing the effect size or test type affect the required sample size?

Changing the effect size or test type directly impacts the required sample size. For example, increasing the effect size reduces the sample size needed to detect the effect, while using a one-tailed test also reduces the sample size compared to a two-tailed test.

What happens when the test is changed from an independent samples t-test to a paired samples t-test?

When the test is changed to a paired samples t-test, which compares two dependent groups, the required sample size decreases significantly. For example, the total sample size reduces from 128 participants in the independent samples t-test to 12 participants in the paired samples t-test.

What would happen if a study used a one-tailed test instead of a two-tailed test?

If a study used a one-tailed test, the required sample size would be smaller because the test is directional, meaning the effect is expected in only one direction. In the video, the sample size dropped to 42 participants when switching to a one-tailed test.