Agent Skills
› aiming-lab/AutoResearchClaw
› statistical-reporting
statistical-reporting
GitHub提供统计检验选择、假设检查及APA格式报告的最佳实践，辅助实验结果分析与论文结果章节撰写。
Trigger Scenarios
分析实验数据并选择合适的统计检验方法
检查统计假设如正态性和方差齐性
按照APA规范撰写和格式化统计报告
Install
npx skills add aiming-lab/AutoResearchClaw --skill statistical-reporting -g -y
SKILL.md
Frontmatter
{
"name": "statistical-reporting",
"metadata": {
"author": "researchclaw",
"version": "1.0",
"category": "writing",
"priority": "3",
"references": "adapted from K-Dense-AI\/claude-scientific-skills",
"trigger-keywords": "statistic,hypothesis test,p-value,regression,ANOVA,t-test,effect size,confidence interval",
"applicable-stages": "14,17"
},
"description": "Statistical test selection, assumption checking, and APA-formatted reporting. Use when analyzing experimental results or writing results sections."
}
Statistical Reporting Best Practice
Test Selection Quick Reference
- Comparing two groups (independent, normal): Independent t-test
- Comparing two groups (independent, non-normal): Mann-Whitney U test
- Comparing two groups (paired, normal): Paired t-test
- Comparing two groups (paired, non-normal): Wilcoxon signed-rank test
- Comparing 3+ groups (independent, normal): One-way ANOVA + post-hoc
- Comparing 3+ groups (non-normal): Kruskal-Wallis test
- Relationship between continuous variables: Pearson or Spearman correlation
- Categorical outcomes: Chi-square or Fisher's exact test
- Predicting continuous outcome: Linear regression
- Predicting binary outcome: Logistic regression
Assumption Checking
- Normality: Shapiro-Wilk test (n < 50) or visual Q-Q plots
- Homogeneity of variance: Levene's test before t-tests and ANOVA
- Independence: Verify study design ensures independent observations
- Linearity: Scatter plots and residual plots for regression
- Multicollinearity: VIF < 5 for multiple regression predictors
- When assumptions are violated, use non-parametric alternatives or robust methods
APA Reporting Format
- t-test: t(df) = X.XX, p = .XXX, d = X.XX
- ANOVA: F(df_between, df_within) = X.XX, p = .XXX, eta-squared = .XX
- Correlation: r(df) = .XX, p = .XXX [95% CI: .XX, .XX]
- Chi-square: chi-square(df, N = XXX) = X.XX, p = .XXX
- Regression: beta = X.XX, SE = X.XX, t = X.XX, p = .XXX
- Always report exact p-values (not "p < .05") unless p < .001
- Use leading zero for values that can exceed 1 (e.g., t = 0.50) but not for those bounded by 1 (e.g., p = .032, r = .45)
Effect Sizes
- ALWAYS report effect sizes alongside p-values
- Cohen's d for group comparisons: small = 0.2, medium = 0.5, large = 0.8
- Eta-squared for ANOVA: small = .01, medium = .06, large = .14
- R-squared for regression: report adjusted R-squared for multiple predictors
- Odds ratios for logistic regression with 95% confidence intervals
- Distinguish statistical significance from practical significance
Common Mistakes to Avoid
- Never say "the results were not significant, therefore there is no effect"
- Do not confuse correlation with causation in observational data
- Apply multiple comparison corrections (Bonferroni, FDR) when running many tests
- Report confidence intervals, not just point estimates
- State whether tests are one-tailed or two-tailed and justify the choice
Version History
- e2e23c9 Current 2026-07-25 07:47
