数据分析
statistics
# Statistical Analysis & Quality Control
文件预览
1 个文件
SKILL.md
2.2 KB · 可预览
--- name: statistics description: # Statistical Analysis & Quality Control --- # Statistical Analysis & Quality Control ## Overview Comprehensive statistical methodology for scientific research. Covers test selection, assumption verification, power analysis, effect size reporting, and reporting standards. ## Test Selection Guide | Data Type | Groups | Paired? | Normal? | Recommended Test | |-----------|--------|---------|---------|-----------------| | Continuous | 2 | No | Yes | Independent t-test | | Continuous | 2 | No | No | Mann-Whitney U | | Continuous | 2 | Yes | Yes | Paired t-test | | Continuous | 2 | Yes | No | Wilcoxon signed-rank | | Continuous | 3+ | No | Yes | One-way ANOVA + post hoc | | Continuous | 3+ | No | No | Kruskal-Wallis + Dunn | | Continuous | 3+ | Yes | Yes | Repeated measures ANOVA | | Categorical | 2x2 | — | — | Chi-square / Fisher's exact | | Time-to-event | 2+ | — | — | Log-rank + KM curves | | Time-to-event | Adjusted | — | — | Cox proportional hazards | | Continuous | Prediction | — | — | Linear/logistic regression | ## Assumption Checks - **Normality**: Shapiro-Wilk (n < 50), Kolmogorov-Smirnov (n > 50), Q-Q plot visual - **Homoscedasticity**: Levene's test, Bartlett's test - **Independence**: study design review (not a statistical test) - **Proportional hazards**: Schoenfeld residuals, log-log plot ## Multiple Comparison Correction - **Bonferroni**: conservative, for few comparisons - **Holm-Bonferroni**: step-down, less conservative than Bonferroni - **FDR (Benjamini-Hochberg)**: for many comparisons (e.g., genomics) - **Tukey HSD**: for all pairwise comparisons after ANOVA ## Effect Size Guidelines | Measure | Small | Medium | Large | |---------|-------|--------|-------| | Cohen's d | 0.2 | 0.5 | 0.8 | | Pearson r | 0.1 | 0.3 | 0.5 | | Odds Ratio | 1.5 | 2.5 | 4.3 | | R-squared | 0.02 | 0.13 | 0.26 | ## Reporting Standards - Always report: test statistic, degrees of freedom, exact p-value, effect size, 95% CI - Follow STROBE (observational), CONSORT (RCTs), PRISMA (reviews), ARRIVE (animal) - Never say "trend toward significance" for p > 0.05 - Report non-significant results honestly
SKILL.md
元数据
| name | statistics |
|---|---|
| description | # Statistical Analysis & Quality Control |
Statistical Analysis & Quality Control
Overview
Comprehensive statistical methodology for scientific research. Covers test selection, assumption verification, power analysis, effect size reporting, and reporting standards.
Test Selection Guide
| Data Type | Groups | Paired? | Normal? | Recommended Test |
|---|---|---|---|---|
| Continuous | 2 | No | Yes | Independent t-test |
| Continuous | 2 | No | No | Mann-Whitney U |
| Continuous | 2 | Yes | Yes | Paired t-test |
| Continuous | 2 | Yes | No | Wilcoxon signed-rank |
| Continuous | 3+ | No | Yes | One-way ANOVA + post hoc |
| Continuous | 3+ | No | No | Kruskal-Wallis + Dunn |
| Continuous | 3+ | Yes | Yes | Repeated measures ANOVA |
| Categorical | 2x2 | — | — | Chi-square / Fisher's exact |
| Time-to-event | 2+ | — | — | Log-rank + KM curves |
| Time-to-event | Adjusted | — | — | Cox proportional hazards |
| Continuous | Prediction | — | — | Linear/logistic regression |
Assumption Checks
- Normality: Shapiro-Wilk (n < 50), Kolmogorov-Smirnov (n > 50), Q-Q plot visual
- Homoscedasticity: Levene's test, Bartlett's test
- Independence: study design review (not a statistical test)
- Proportional hazards: Schoenfeld residuals, log-log plot
Multiple Comparison Correction
- Bonferroni: conservative, for few comparisons
- Holm-Bonferroni: step-down, less conservative than Bonferroni
- FDR (Benjamini-Hochberg): for many comparisons (e.g., genomics)
- Tukey HSD: for all pairwise comparisons after ANOVA
Effect Size Guidelines
| Measure | Small | Medium | Large |
|---|---|---|---|
| Cohen's d | 0.2 | 0.5 | 0.8 |
| Pearson r | 0.1 | 0.3 | 0.5 |
| Odds Ratio | 1.5 | 2.5 | 4.3 |
| R-squared | 0.02 | 0.13 | 0.26 |
Reporting Standards
- Always report: test statistic, degrees of freedom, exact p-value, effect size, 95% CI
- Follow STROBE (observational), CONSORT (RCTs), PRISMA (reviews), ARRIVE (animal)
- Never say "trend toward significance" for p > 0.05
- Report non-significant results honestly