The Clinical Sample Size Calculator computes any one of {α, power, effect size, sample size} when the other three are provided. It supports six statistical designs (one-sample t / two-sample t / paired t / one-way ANOVA / chi² two-proportion / McNemar paired / log-rank survival) and three hypothesis types (superiority / non-inferiority / equivalence TOST). The engine is 100% self-implemented with zero external statistics libraries (Acklam 2003 inverse normal + Numerical Recipes 2e §6.4 Lentz CF incomplete beta + Lenth 1989 non-central t + Poisson-weighted non-central F/χ² direct summation — Patnaik 1949 approximation excluded per Cross-Check Report §1.2), validated against G*Power Manual T1-T5 (n=64 / N=180 / N=108 / N≈117 / events=247).

(1) Alternative to G*Power (Windows/Mac desktop, English-only), PASS NCSS (annual six-figure license), and nQuery (CRO license) — instant web + Korean + mobile (2) Low barrier to entry for clinical statistics / CRA / public-health MS/PhD students (no more R `pwr` coding overhead) (3) Built-in FDA / EMA / MFDS regulatory baselines (Context-Aware Smart Defaults: Superiority two-sided α=0.05 / NI one-sided α=0.025 auto-locked with Sides disabled / Equivalence TOST one-sided α=0.05 parallel on both sides) (4) Auto-generated CONSORT 2010 §7a sample size justification English template — copy directly into IRB / IND submissions

① Pick design (6 cards: one-sample / two-sample / paired t / ANOVA / chi² / McNemar / log-rank) ② Pick hypothesis (Superiority / Non-inferiority / Equivalence TOST) • NI auto-enforces α=0.025 one-sided + Sides disabled lock (FDA 2016 / MFDS 2018 guidance) • Equivalence enforces one-sided α=0.05 on each side (95% CI = two-sided α=0.10) with auto info panel ③ Enter effect size — Split-Screen Workspace: • Left: Direct Parameter (d / f / w / h / OR / HR) + Cohen 1988 S/M/L benchmarks • Right: Raw Helper (μ₁/μ₂/σ → d / p₁/p₂ → h / OR → d / HR → ln HR) — real-time bidirectional sync (no apply button) ④ Enter α / power / allocation ratio / number of multiple tests ⑤ Auto-computes: n per group / total N / achieved power / critical value + design-specific (ANOVA k / chi² df / McNemar ψ·p_d / log-rank HR·P_event·dropout) ⑥ Visualizations: Power Curve (custom SVG, 80% target dashed line + current N marker) + **2D Sensitivity Matrix Heatmap** (5×5 grid, Power ≥ 80% Regulatory Compliant Zone emerald border) ⑦ Export: PNG (heatmap) / CSV (input + result + 25 cells) / JSON (full state) + **One-click copy of CONSORT 2010 §7a sample size justification English template**

Example 1) Two-sample t-test (G*Power Manual §3 / §11.2) → Cohen's d = 0.5, α = 0.05 (two-sided), power = 0.80, 1:1 → **n_per_group = 64, total N = 128** Example 2) One-way ANOVA (G*Power Manual §4 / §10.3) → Cohen's f = 0.25, α = 0.05, power = 0.80, k = 4 groups → **N = 180** Example 3) Chi² two-proportion (Cohen 1988 Table 7.4.4) → Cohen's w = 0.3, df = 1, α = 0.05, power = 0.80 → **N = 108** Example 4) McNemar paired (Connor 1987 Table 1) → Odds Ratio ψ = 2.0, p_discordant = 0.30, α = 0.05, power = 0.80 → **N ≈ 117 pairs** Example 5) Schoenfeld log-rank (Schoenfeld 1981 / CWS §7.2) → HR = 0.7, α = 0.05 (two-sided), power = 0.80, 1:1 → **events d = 247** → P_event = 0.7 → N ≈ 353; 10% dropout adjustment → N ≈ 393 Example 6) Non-inferiority (FDA NI 2016 §IV.A) → Switch hypothesis to NI → α = 0.025 one-sided auto-enforced → info panel shown