KONSİLCalculators · Comparison
G*Power alternative: sample size calculators in your browser, no installation
G*Power is a free power analysis program developed by Buchner, Erdfelder, Faul and Lang and distributed from Heinrich Heine University Düsseldorf; you install it on your computer. KONSİL’s calculators cover ten sample size methods in the browser, with no installation or sign-up. This page explains what differs, when each one fits better, and how KONSİL’s results line up with published G*Power examples.
Side-by-side comparison
| Feature | G*Power | KONSİL calculators |
|---|---|---|
| How it runs | A program installed on your computer | In the browser; no installation or sign-up |
| Operating systems | Windows (3.1.9.7) and macOS (3.1.9.6, built for Intel processors; a native Apple silicon version, G*Power 4, is in development). The official page does not mention Linux. | Any device with a browser |
| Cost | Free for everyone; commercial distribution is prohibited | Free |
| Coverage | t, F, χ², z and some exact tests; the manual also documents regression, correlation, logistic and Poisson regression, Wilcoxon tests, McNemar and Fisher’s exact test | Ten methods (listed below) |
| Types of analysis | A priori, compromise, criterion, post hoc and sensitivity | A priori sample size; for some methods, the smallest effect detectable with the number of participants you can reach |
| Output | Calculation window and plots | The result, an attrition table and a methods sentence you can paste into a protocol |
| Citation | Faul et al. (2007, 2009) | The sources listed at the end of each calculator page |
The G*Power details come from the program’s official page and manual; version numbers and operating system support may change over time.
When G*Power is the better choice
KONSİL’s ten calculators cover only part of what G*Power covers. G*Power fits better when:
- The test you need is not among our calculators. The manual also documents procedures such as regression, correlation, logistic and Poisson regression, the Wilcoxon signed-rank test, the test of equality of two variances and Fisher’s exact test.
- You need a type of analysis beyond a priori: G*Power also offers compromise, criterion, post hoc and sensitivity analyses.
- You want an exact method. KONSİL’s McNemar calculator uses the Connor (1987) approximation; G*Power offers an exact test and a fast approximation for McNemar.
- A journal, committee or supervisor expects G*Power to be cited in the methods section.
When the KONSİL calculators are enough
When you need a quick calculation from any device, with no installation or sign-up, and explanations in Turkish or English; it works on a phone too. Each calculator gives the attrition table and a methods sentence you can paste into a protocol next to the result, and the formula basis and sources are listed at the end of every page.
The equivalent in G*Power
| KONSİL calculator | Procedure in the G*Power manual | Manual chapter |
|---|---|---|
| Two-sample t test | t test: Means - difference between two independent means (two groups) | 21 |
| Paired t test | t test: Means - difference between two dependent means (matched pairs) | 19 |
| One-way ANOVA | F test: Fixed effects ANOVA - one way | 10 |
| Mann–Whitney | Wilcoxon-Mann-Whitney test of a difference between two independent means | 24 |
| McNemar (paired) | Exact: Proportion - inequality, two dependent groups (McNemar) | 5 |
No counterpart is given in this table for the other calculators (survey and prevalence, two proportions, non-inferiority of means, sensitivity and specificity precision, sensitivity threshold test).
Do the results match G*Power?
The three examples below come from G*Power’s own publications; with the same inputs KONSİL gives the same number.
| Input | Published G*Power result | KONSİL |
|---|---|---|
| Two independent groups t test, one-tailed, d = 0.5, α = 0.05, power = 95%, equal groups | Total 176 (88 per group). Source: Faul et al. (2007) | Total 176 (88 per group) |
| One-sample t test, one-tailed, d = 0.625, α = 0.05, power = 95% | N = 30. Source: G*Power manual, chapter 20.3 | 30 (paired t test calculator, d_z = 0.625) |
| One-way ANOVA, 10 groups, Cohen’s f = 0.25, α = 0.05, power = 95%, equal groups | Total 390 (39 per group); actual power 0.952363. Source: G*Power manual, chapter 10.3 | Total 390 (39 per group); actual power 0.9524 |
The α option in the calculators is two-sided. Because the first two examples are one-tailed tests, choosing two-sided α = 0.10 in KONSİL reproduces the same numbers. With two-sided α = 0.05 and the same inputs the numbers are larger: 210 in total for the first example and 36 pairs for the second. The third example is an F test, which has no one- or two-sided option; α = 0.05 is entered as is.
This comparison rests on three published examples; it is not a comparison of every G*Power output. KONSİL’s power solver is also tested against an independent implementation (Python statsmodels).
The ten calculators
- Two-sample t-test sample size calculator
- Paired t-test sample size calculator
- Sample size calculator for comparing two proportions
- One-way ANOVA sample size calculator
- Mann-Whitney U test sample size calculator
- Non-inferiority sample size calculator (means)
- Sensitivity / specificity sample size calculator
- Sample size: sensitivity above a threshold
- McNemar sample size calculator (paired tests)
- Survey sample size calculator (Cochran, Yamane)
What to write in the methods section
Name the tool you used. For G*Power, cite Faul et al. (2007); for a KONSİL calculator, copy the methods sentence the page gives and cite the sources listed on the page. In both cases state the inputs (effect size, α, power) and what the effect size is based on.
Sources
- Faul F, Erdfelder E, Lang A-G, Buchner A. G*Power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behav Res Methods. 2007;39(2):175–191.
- G*Power 3.1 manual, 1 June 2023. Heinrich-Heine-Universität Düsseldorf (PDF).
- G*Power. Heinrich-Heine-Universität Düsseldorf (the program's official page).
- Connor RJ. Sample size for testing differences in proportions for the paired-sample design. Biometrics. 1987;43:207–211.