Practical examples of G*Power power analysis examples for real studies

Why start with real examples of G*Power power analysis examples

Statistical power is not just a theoretical nicety; it’s the difference between a study that can actually detect a meaningful effect and one that wastes everyone’s time. G*Power is popular because it’s free, flexible, and relatively friendly once you’ve seen it used in context. The best examples are always grounded in real questions:

• Will this new treatment improve outcomes enough to matter?
• Is this teaching method actually better than the status quo?
• Does this predictor (like BMI or hours of sleep) explain a meaningful share of outcome variance?

The sections below walk through several examples of G*Power power analysis examples that mirror what you’d see in health research, psychology, education, and social science. I’ll stick to common study designs you can reproduce directly in the software.

Example of G*Power power analysis: two-group clinical trial (independent t-test)

Imagine you’re planning a randomized clinical trial comparing a new blood-pressure medication to a standard drug. Your main outcome is systolic blood pressure (SBP) after 12 weeks.

Research question
Does the new drug reduce SBP more than the standard drug?

Design
Two independent groups (new vs standard), continuous outcome → independent-samples t-test.

Key inputs for GPower
In GPower, you’d choose:

• Test family: t tests
• Statistical test: Means: Difference between two independent means (two groups)
• Type of power analysis: A priori: Compute required sample size

Now the parameters:

• Tail(s): two-tailed (you allow for the possibility the new drug could be better or worse)
• Significance level (α): 0.05 (standard in many clinical studies; see NIH guidance on conventional thresholds at nih.gov)
• Desired power (1 − β): 0.80 (often considered a reasonable minimum in practice)
• Effect size d: 0.5 (a moderate effect; based on prior pilot data or literature)

In G*Power, with these inputs, you’d get a required sample size of about 64 participants per group (128 total) to achieve 80% power to detect a moderate effect.

Why this matters in 2024–2025
Clinical trials are under increasing pressure from regulators and ethics boards to justify sample sizes. Underpowered trials mean unnecessary exposure of participants to risk with little chance of a clear result. This kind of example of G*Power power analysis is exactly what appears in many trial protocols and IRB submissions.

Education study: one-way ANOVA with three teaching methods

Now switch to an education context. You’re evaluating three different methods for teaching introductory statistics: traditional lecture, flipped classroom, and fully online.

Research question
Do average final exam scores differ across the three teaching methods?

Design
Three independent groups, continuous outcome → one-way ANOVA.

G*Power setup

• Test family: F tests
• Statistical test: ANOVA: Fixed effects, omnibus, one-way
• Type of power analysis: A priori

Parameters:

• α = 0.05
• Power = 0.80
• Number of groups = 3
• Effect size f = 0.25 (medium; based on Cohen’s conventions and perhaps previous meta-analyses on teaching interventions)

In G*Power, this yields a total sample size of about 159 students (roughly 53 per teaching method) to detect a medium-sized difference across the three groups.

Real-world context
Many universities now compare in-person vs hybrid vs online formats. To make a credible claim that one format outperforms the others, you need a power analysis like this. As more institutions move to data-informed teaching decisions, these kinds of examples of G*Power power analysis examples are increasingly common in education research published by universities and teaching centers.

Repeated-measures psychology experiment: within-subject t-test

Psychology studies often use repeated-measures designs, which are statistically efficient because each participant serves as their own control.

Scenario
You’re testing whether a mindfulness app reduces self-reported anxiety scores after four weeks.

Design
Participants complete an anxiety scale before and after using the app. Same participants, two time points → paired-samples t-test.

G*Power configuration

• Test family: t tests
• Statistical test: Means: Difference between two dependent means (matched pairs)
• Type of power analysis: A priori

Inputs:

• α = 0.05
• Power = 0.90 (you want higher power because you expect a moderate-to-large effect and the study is relatively cheap to run)
• Effect size dz = 0.5 (moderate within-subject effect)

G*Power will output a required sample size of about 44 participants.

Why this is one of the best examples for students
This is one of the best examples of G*Power power analysis examples to teach because it highlights how repeated-measures designs need fewer participants than independent-group designs for the same effect size. That’s a powerful lesson for anyone planning lab-based psychology or behavioral studies.

Public health example: multiple linear regression

Let’s move to a regression setting that’s common in epidemiology and public health.

Scenario
You want to model systolic blood pressure as a function of age, BMI, physical activity, and smoking status in adults. Your main interest is whether physical activity is a meaningful predictor after controlling for others.

Design
Multiple linear regression with 4 predictors.

G*Power steps

• Test family: F tests
• Statistical test: Linear multiple regression: Fixed model, R² deviation from zero
• Type of power analysis: A priori

Inputs:

• α = 0.05
• Power = 0.80
• Number of predictors = 4
• Effect size f² = 0.10 (small-to-moderate; corresponds to R² ≈ 0.09)

G*Power will return a required total sample size of about 118 participants.

Context in 2024–2025
Public health studies often rely on observational data from cohorts or surveys (for example, NHANES data described at cdc.gov). When designing new cohorts or sub-studies, investigators frequently use examples of G*Power power analysis examples like this to justify that they can detect meaningful R² values or incremental variance explained by a key predictor.

Logistic regression example: predicting disease status

Many modern datasets involve binary outcomes: disease vs no disease, dropout vs completion, relapse vs no relapse.

Scenario
You’re studying whether a new screening questionnaire predicts presence of type 2 diabetes (yes/no) in adults.

Design
Binary outcome, continuous predictor → logistic regression.

GPower does not directly support every logistic regression nuance, but you can approximate using tests on proportions or use the *z tests → Logistic regression module in newer versions.

Approximate G*Power setup (two proportions)

Suppose you categorize people as “high-risk” vs “low-risk” based on the questionnaire and compare diabetes prevalence.

• Test family: z tests
• Statistical test: Proportions: Inequality, two independent groups (unconstrained)
• Type of power analysis: A priori

Inputs:

• α = 0.05
• Power = 0.80
• Proportion in control (low-risk) group: 0.10 (10% diabetes prevalence)
• Proportion in high-risk group: 0.20 (you expect double the prevalence)
• Allocation ratio N2/N1 = 1

G*Power will suggest a total sample size of roughly 400 participants (about 200 per group) to detect this difference with 80% power.

Why this matters
Screening tools are widely evaluated in modern preventive medicine. Power analyses like this, even when approximate, show whether a study can detect a clinically meaningful difference in risk. These real examples of G*Power power analysis examples are frequently seen in early-phase diagnostic and screening research.

Repeated-measures ANOVA: three time points in a longitudinal study

Longitudinal designs are everywhere in 2024–2025, from mental health apps to remote patient monitoring.

Scenario
You track depression scores at baseline, 3 months, and 6 months after starting therapy.

Design
Same participants measured three times → repeated-measures ANOVA, within factors.

G*Power setup

• Test family: F tests
• Statistical test: ANOVA: Repeated measures, within factors
• Type of power analysis: A priori

Inputs (simplified example):

• α = 0.05
• Power = 0.80
• Number of measurements = 3
• Effect size f = 0.25 (medium)
• Correlation among repeated measures: 0.5 (reasonable for psychological scales)
• Nonsphericity correction ε = 1 (idealized; in practice you may set this lower if you expect violations)

G*Power will estimate that you need about 28–30 participants to detect a medium within-subject effect across time.

Practical note
Because attrition is a real issue in longitudinal work, many researchers inflate this number by 20–30% in planning. This is another example of G*Power power analysis where the output is a starting point, not the final word.

Factorial ANOVA example: 2×2 design in social psychology

Factorial designs are common when you want to test both main effects and interactions.

Scenario
You test whether message framing (gain vs loss) and source credibility (expert vs peer) affect vaccination intention scores.

Design
A 2×2 between-subjects factorial ANOVA (4 groups total).

G*Power configuration

• Test family: F tests
• Statistical test: ANOVA: Fixed effects, special, main effects and interactions
• Type of power analysis: A priori

Inputs:

• α = 0.05
• Power = 0.80
• Number of groups = 4
• Effect size f = 0.20 for the interaction (you expect a small-to-medium interaction)

G*Power will return a total sample size around 200 participants (about 50 per cell) to detect a small-to-medium interaction effect with adequate power.

Why this is a strong teaching example
Students often underestimate how many participants are needed to detect interactions. This example of G*Power power analysis shows that chasing subtle interaction effects is expensive in terms of sample size.

Common mistakes when using these examples of G*Power power analysis examples

Even with these real examples of G*Power power analysis examples, there are recurring pitfalls:

1. Using effect sizes that are too optimistic
If you always plug in large effect sizes (d = 0.8, f = 0.40) because they give you smaller sample sizes, you’re setting yourself up for underpowered studies. A better approach is to:

• Look up effect sizes in meta-analyses or large studies.
• Use conservative estimates when evidence is uncertain.
• Consider pilot data, but remember small pilots often exaggerate effects.

2. Ignoring multiple outcomes and tests
If you plan many tests (multiple outcomes, subgroups, or time points), the nominal α = 0.05 assumption may be too generous. Some teams adjust α (for example, to 0.01) or focus the power analysis on the primary endpoint only, as recommended in many NIH and FDA guidelines.

3. Treating GPower output as a guarantee
A power analysis is only as good as its assumptions. Violations of normality, unequal variances, missing data, and measurement error can all reduce actual power. This is why the best examples of GPower power analysis examples are always paired with a data analysis plan that anticipates these issues.

4. Forgetting practical constraints
G*Power might say you need 500 participants, but your clinic only sees 150 eligible patients a year. In that case, you might:

• Accept lower power and acknowledge it explicitly.
• Extend recruitment across sites.
• Simplify the design to reduce the required sample size.

FAQ: Short answers built around real examples

Q1. Where can I find more examples of GPower power analysis examples for my field?
Look at published articles in journals in your area and read the Methods sections carefully. Many psychology and education journals now require a power analysis statement. For medical and health research, NIH-funded trials often include GPower-based justifications in their protocols, and you can see summaries in ClinicalTrials.gov entries or related papers.

Q2. Can I use one example of GPower power analysis and just copy the numbers?
Not safely. These examples include effect sizes, α levels, and power targets that may not match your question, outcome variability, or ethical constraints. You can use the structure of these examples of GPower power analysis examples (choice of test, input fields) but you should adapt the effect size and parameters based on your own context and literature.

Q3. How do I choose a realistic effect size for my G*Power analysis?
Start with prior research: meta-analyses, large cohort studies, or clinical guidelines. For health topics, organizations like the National Institutes of Health and Mayo Clinic often summarize treatment effects or risk differences. If you have no prior data, consider running a small pilot and then using a slightly smaller effect size than the pilot suggests to avoid overestimating.

Q4. Are these examples of GPower power analysis examples acceptable for IRB or ethics review?
Yes, as long as they are tailored to your study and justified with references. Ethics committees typically expect: the test type, α level, target power, effect size (with justification), and resulting sample size. Many reviewers are familiar with GPower outputs.

Q5. Can GPower handle more advanced models like mixed-effects or multilevel models?
Not directly in all cases. GPower is strongest for classic designs: t-tests, ANOVA/ANCOVA, correlation, basic regression, and some repeated-measures setups. For complex multilevel models, you may need specialized tools (for example, simulation-based power using R packages). Still, examples of G*Power power analysis examples are a good starting point when your design can be reasonably approximated by simpler tests.

If you treat these scenarios as templates rather than copy-and-paste recipes, you’ll be able to build your own real examples of G*Power power analysis examples that satisfy reviewers, protect participants, and give your study a fighting chance of answering the question you care about.