Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Medical biostatistics hypothesis testing toolkit: t-tests, ANOVA, chi-square, Fisher exact, Mann-Whitney, Kruskal-Wallis, sample size calculation, power analysis, multiple testing correction, survival analysis, and clinical trial biostatistics.
.claude/skills/mkurman-biostatistics/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | 12% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 25% | 0% |
Medical biostatistics for hypothesis testing, clinical trial design, and survival analysis. Covers t-tests, ANOVA, chi-square, Fisher exact, Mann-Whitney, Kruskal-Wallis, sample size calculation, power analysis, and multiple testing correction.
bashuv pip install scipy statsmodels
pythonimport numpy as np from scipy import stats # Two-sample t-test t_stat, p = stats.ttest_ind(np.random.normal(100, 15, 30), np.random.normal(110, 15, 30)) # Mann-Whitney u_stat, p = stats.mannwhitneyu(np.random.normal(100, 15, 30), np.random.normal(110, 15, 30)) # Chi-square chi2, p, dof, _ = stats.chi2_contingency(np.array([[30, 10], [20, 40]]))
pythonfrom statsmodels.stats.power import TTestIndPower n = TTestIndPower().solve_power(effect_size=0.5, power=0.80, alpha=0.05) print(f"N per group: {np.ceil(n):.0f}")
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 12,688 | 12,058 | -5% | 1 | 1 | 0% | 2,508 | 2,980 | +19% | 0 | 0 | — |
case-02 | fail→pass | 5,013 | 2,121 | -58% | 1 | 1 | 0% | 801 | 665 | -17% | 0 | 0 | — |
case-03 | pass→pass | 6,756 | 5,819 | -14% | 1 | 1 | 0% | 1,387 | 1,548 | +12% | 0 | 0 | — |
case-04 | pass→pass | 7,409 | 7,294 | -2% | 1 | 1 | 0% | 1,501 | 1,870 | +25% | 0 | 0 | — |
case-05 | pass→pass | 10,758 | 7,640 | -29% | 1 | 1 | 0% | 2,184 | 1,872 | -14% | 0 | 0 | — |
case-06 | pass→pass | 6,183 | 3,533 | -43% | 1 | 1 | 0% | 1,212 | 1,008 | -17% | 0 | 0 | — |
case-07 | pass→pass | 9,909 | 6,758 | -32% | 1 | 1 | 0% | 1,782 | 1,565 | -12% | 0 | 0 | — |
case-08 | pass→pass | 4,237 | 2,259 | -47% | 1 | 1 | 0% | 738 | 731 | -1% | 0 | 0 | — |
case-09 | fail→pass | 6,719 | 4,077 | -39% | 1 | 1 | 0% | 1,097 | 1,065 | -3% | 0 | 0 | — |
case-10 | pass→pass | 5,037 | 2,916 | -42% | 1 | 1 | 0% | 954 | 809 | -15% | 0 | 0 | — |
case-11 | pass→pass | 4,784 | 3,584 | -25% | 1 | 1 | 0% | 821 | 927 | +13% | 0 | 0 | — |
case-12 | pass→pass | 4,842 | 2,173 | -55% | 1 | 1 | 0% | 786 | 719 | -9% | 0 | 0 | — |
case-13 | pass→pass | 5,533 | 1,893 | -66% | 1 | 1 | 0% | 976 | 675 | -31% | 0 | 0 | — |
case-14 | pass→pass | 4,266 | 3,891 | -9% | 1 | 1 | 0% | 720 | 1,023 | +42% | 0 | 0 | — |
case-15 | pass→pass | 5,839 | 3,103 | -47% | 1 | 1 | 0% | 1,058 | 927 | -12% | 0 | 0 | — |
case-16 | pass→pass | 4,533 | 3,500 | -23% | 1 | 1 | 0% | 850 | 991 | +17% | 0 | 0 | — |
case-17 | pass→pass | 8,710 | 6,234 | -28% | 1 | 1 | 0% | 1,670 | 1,504 | -10% | 0 | 0 | — |
case-18 | pass→pass | 8,895 | 5,163 | -42% | 1 | 1 | 0% | 1,639 | 1,377 | -16% | 0 | 0 | — |
case-19 | fail→pass | 5,134 | 2,873 | -44% | 1 | 1 | 0% | 846 | 887 | +5% | 0 | 0 | — |
case-20 | pass→pass | 9,144 | 5,463 | -40% | 1 | 1 | 0% | 1,712 | 1,376 | -20% | 0 | 0 | — |
case-21 | pass→pass | 11,763 | 11,134 | -5% | 1 | 1 | 0% | 1,908 | 2,335 | +22% | 0 | 0 | — |
case-22 | pass→pass | 11,149 | 10,659 | -4% | 1 | 1 | 0% | 1,812 | 2,187 | +21% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +14 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.