2026-10-07
How to calculate sample size for your thesis (G*Power explained simply)
"How many patients do I need?" is the first question every committee asks. A sample that is too small cannot detect a real effect; a sample that is too large wastes time and patients. The answer comes from a power analysis, usually done in the free program G*Power.
The four numbers you need
Every sample size calculation needs four things:
- The test you will use (e.g. comparing two groups with an independent t-test).
- Alpha (α) — the risk of a false positive. Almost always 0.05.
- Power (1 − β) — the chance of detecting a real effect. Usually 0.80 (80%).
- Effect size — how big the difference you expect is. This is the most important and the most difficult number.
Where does the effect size come from?
The best source is a previous study similar to yours. For two groups, Cohen's d is:
d = (mean of group 1 − mean of group 2) ÷ pooled standard deviation
If a previous trial found pain scores of 3.2 and 4.4 with a standard deviation of about 2.0, then d = 1.2 ÷ 2.0 = 0.6.
If no study exists, use a pilot study, or Cohen's conventions: small 0.2, medium 0.5, large 0.8. Always state in your Methods where your effect size came from — supervisors check this.
Example 1: two independent groups (RCT)
In G*Power choose:
- Test family: t tests
- Statistical test: Means: Difference between two independent means (two groups)
- Type of power analysis: A priori
- Tails: two · Effect size d: 0.5 · α: 0.05 · Power: 0.80 · Allocation ratio: 1
Result: 64 per group, 128 in total. With a large effect (d = 0.8) you need only 26 per group.
Example 2: one group, before and after (pre–post)
Choose Means: Difference between two dependent means (matched pairs). With d = 0.5, α = 0.05 and power = 0.80 you need 34 participants. Paired designs need fewer people because each person is compared with themselves.
Example 3: correlation
Choose test family Exact, then Correlation: Bivariate normal model. To detect r = 0.3 (two-tailed, α = 0.05, power = 0.80) you need about 84 participants.
Do not forget drop-outs
Patients leave studies. If you expect 20% drop-out, divide your number by 0.8: 128 ÷ 0.8 = 160 participants to recruit.
How to write it in your Methods
"The sample size was calculated using G*Power 3.1 for an independent t-test, with an effect size of 0.6 based on [previous study], α = 0.05 and power = 0.80. The required sample was 45 per group; to allow for 20% drop-out, 57 participants per group were recruited."
Calculate it in one minute with Neura
Neura's Sample size tool covers every common thesis design (two groups, pre–post, ANOVA, correlation, proportions, regression and more). Its results match G*Power, it shows a power curve, and it writes the paragraph for your Methods chapter with the reference. The first calculation every day is free.
Summary: choose the test → α 0.05 → power 0.80 → effect size from a previous study → add drop-out → write it clearly in your Methods.