Sample Size Calculator (G*Power)
A-priori power analysis for multiple regression — more defensible than population-only formulas.
1 Set your parameters
* Every independent variable in the model, including controls and dummy codes.
💡 Why power analysis?
Population-based formulas such as Taro Yamane ignore how many relationships your model has to estimate. A power analysis confirms your sample is large enough to actually detect the effects your hypotheses predict — which is what reviewers ask about.
Minimum sample size
For 5 predictors at 80% power, α = 0.05
Methodology paragraph
ⓘ This free tool helps with a first calculation. Khontob does not guarantee the results or any outcome of using them. Before using them in your research, please check them against the original software or textbook and with your advisor.
G*Power sample size table for multiple regression
Minimum total sample size calculated with the same method as G*Power 3.1 (F test, Linear multiple regression: Fixed model, R² deviation from zero) at α = .05 and power = .80. Use it as a first reference and confirm in G*Power before relying on it.
| Predictors (k) | f² = .02 (small) | f² = .15 (medium) | f² = .35 (large) |
|---|---|---|---|
| 1 | 395 | 55 | 25 |
| 2 | 485 | 68 | 31 |
| 3 | 550 | 77 | 36 |
| 4 | 602 | 85 | 40 |
| 5 | 647 | 92 | 43 |
| 6 | 688 | 98 | 46 |
| 7 | 725 | 103 | 49 |
| 8 | 759 | 109 | 52 |
| 10 | 822 | 118 | 57 |
| 12 | 878 | 127 | 61 |
| 15 | 954 | 139 | 68 |
| 20 | 1,064 | 157 | 77 |
How to read it: with 5 predictors and a medium effect you need at least 92 respondents. For other α or power values, use the calculator above.
Choosing your parameters
1. Counting predictors
Count every independent variable that enters the model. If your framework has four personal factors and four marketing factors predicting purchase intention, that is eight predictors — not two constructs. Control variables and dummy-coded categories count too.
Weak expected relationships — requires a much larger sample.
The default convention when no prior estimate exists (recommended).
Only when prior evidence supports a strong effect.
! Collect more than the minimum
The computed figure is a statistical minimum. Plan to over-collect by 10–20% to absorb incomplete questionnaires, straight-lined responses, and outliers removed during cleaning.
Reference: Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum. · Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009). Statistical power analyses using G*Power 3.1. Behavior Research Methods, 41(4), 1149–1160.
Frequently asked questions
Why use G*Power instead of the Taro Yamane formula?
The Taro Yamane formula derives sample size from population size alone. G*Power derives it from statistical power, significance level, and effect size, which is the standard for regression-based designs because it accounts for how many predictors the model has to estimate. If your analysis is a multiple regression or an SEM, most supervisors will expect a power-based justification rather than a population-based one.
What effect size (f²) should I use?
If no prior study gives you an estimate, Cohen (1988) recommends the medium convention of f² = 0.15 for multiple regression. Use 0.02 (small) only when you expect weak relationships and can afford a much larger sample, and 0.35 (large) only when prior evidence supports a strong effect.
Is the computed n the final number I should collect?
No. The computed value is the statistical minimum. In practice you should over-collect by 10 to 20 percent to absorb incomplete questionnaires, straight-lined responses, and outliers that get removed during data cleaning.
How do I count the number of predictors?
Count every independent variable that enters the regression model, including control variables and any dummy-coded categories. If your conceptual framework has four personal factors and four marketing factors predicting one outcome, the number of predictors is eight, not two.
Have a website, blog or course page? Add this calculator for free
Copy the code into your page (HTML mode). Works in WordPress, Google Sites and faculty websites — your students can use it without leaving your page.
Let us help with the respondents so you can focus on your research
Tell us about your project. We'll estimate it for free, with no commitment, and reply within 24 hours.