Table 1: Sample Size Calculation by Statistical Test

Test Type Purpose Formula / Method Required Inputs Example Notes / Tools
 
Zα/2 σ 2
One-sample mean Estimate population n = E σ, E, Confidence σ = 6, E = 2, α = Use when comparing
test mean level 0.05 ⇒ n = 35 to known population
mean
 2
Z1−α/2 +Z1−β
Two-sample t-test Compare two means n=2 ∆/σ α, Power, σ, ∆ ∆ = 10, σ = 15, G*Power: Means
(independent) α = 0.05, power=0.8 difference test
⇒ n = 36 per group
 2
(Z1−α/2 +Z1−β )σd
Paired t-test Compare before/after n= ∆d σd , ∆d , α, Power σd = 10, ∆d = 8 ⇒ Higher power due
means n = 25 to within-subject de-
sign
2(Z1−α/2 +Z1−β )2
One-way ANOVA (k Compare ≥ 3 means n= kf 2 α, Power, k, f f = 0.25, k = 3, G*Power: ANOVA

1
groups) α = 0.05, power=0.8 fixed effects
⇒ n ≈ 52 per group
2
Zα/2 p(1−p)
Proportion (1 group) Estimate a population n= E2 p, E, α p = 0.4, E = 0.05 ⇒ Standard for sur-
proportion n = 369 veys/polls
Two-proportion test Compare two propor- n √ √ = p1 , p2 , α, Power p1 = 0.3, p2 = 0.5, ⇒ G*Power: Propor-
tions (Z1−α/2 2p(1−p)+Z1−β p1 (1−p1 )+p2 (1−p2 ))2 n = 88/group tion comparison
(p1 −p2 )2
χ2crit
Chi-square test Association or fit n= w2 w, α, df w = 0.3, df=2 ⇒ n ≈ Use Cohen’s w
88 (0.1–0.5)
(Z1−α/2 +Z1−β )2
Correlation (Pearson Detect linear relation- n= 1+r 2
0.5[ln( 1−r )]
r, α, Power r = 0.3 ⇒ n ≈ 84 G*Power: Correla-
r) ship tion test
(Z1−β +Z1−α/2 )2 (1−R2 )
Simple regression Detect linear effect n= R2 + 2 R2 , α, Power R2 = 0.25 ⇒ n ≈ 30 One predictor only
(Z1−β +Z1−α/2 )2 2 2
Multiple regression Model with several n= f2 +k+1 f , α, Power, k f = 0.15, k = 5 ⇒ Rule: n ≥ 50 + 8k
(k predictors) predictors n ≈ 92
Logistic regression Binary outcome pre- Rule of thumb: 10 Event rate, Predic- 8 predictors, event rate Use R:
diction events/predictor variable tors = 0.3 ⇒ n ≈ 267 powerLogistic
(Z1−α/2 +Z1−β )2
Survival (Cox) Compare survival n= (ln HR)2 ×E HR, Event rate E, HR=0.7, E = 0.6 ⇒ R: powerSurvEpi,
curves α, Power n ≈ 120 events PASS
Factor analysis / Latent structure detec- Empirical: 5–10 subjects per #Items 20 items ⇒ n = More items → larger
PCA tion item 100–200 n
2
Survey (unknown Estimate n = Z p(1−p)
E2 Z, p, E 95% CI, 5% margin ⇒ Apply finite popula-
population) mean/proportion n = 384 tion correction if N
known
Pilot study Estimate σ or p Rule of thumb – 10–30 subjects Used to inform main
 sample size
[Link]
October 2025

1 Introduction