Statistical Analysis Advisor
When to activate
When choosing statistical tests, interpreting results, checking assumptions, or when a reviewer questions your analysis. Use before running any analysis to ensure correct test selection.
When NOT to use
Skip for simple descriptive statistics (mean, SD), data visualization without inference, or when the statistical approach is already validated and approved.
Instructions
Characterize the data:
- Outcome type: Continuous, categorical, ordinal, time-to-event, count
- Groups: Number of groups, paired vs. independent
- Distribution: Normal (Shapiro-Wilk), equal variance (Levene's)
- Sample size per group
Test selection decision tree:
- 2 independent groups, normal → Independent t-test
- 2 independent groups, non-normal → Mann-Whitney U
- 2 paired groups, normal → Paired t-test
- 3+ groups → One-way ANOVA (Kruskal-Wallis if non-normal)
- Repeated measures → RM-ANOVA or Friedman
- Time-to-event → Log-rank test, Cox regression
- Correlation → Pearson (normal) or Spearman (non-normal)
Assumption checks:
- Normality: Shapiro-Wilk (n<50) or Kolmogorov-Smirnov (n≥50)
- Homogeneity of variance: Levene's test
- Sphericity: Mauchly's test (for RM-ANOVA)
- If violated: Report the violation and use the non-parametric alternative
Report results with:
- Test statistic and degrees of freedom
- Exact p-value (not just <0.05)
- Effect size (Cohen's d, η², r, OR as appropriate)
- 95% confidence interval
- Sample size
Multiple comparisons: If >2 comparisons, specify correction:
- Bonferroni (conservative)
- Holm-Bonferroni (step-down)
- Benjamini-Hochberg FDR (exploratory)
Output Format
STATISTICAL ANALYSIS: [Research question]
DATA CHARACTERISTICS:
Outcome: [type] | Groups: [X] | Distribution: [normal/non-normal]
Normality: Shapiro-Wilk W=[X], p=[X]
Variance: Levene's F=[X], p=[X]
TEST SELECTED: [Test name] — [justification]
RESULTS:
[Test]([df]) = [statistic], p = [exact p]
Effect size: [Cohen's d / η² / r] = [value] ([small/medium/large])
95% CI: [[lower], [upper]]
n = [X] per group
INTERPRETATION: [Plain language conclusion]
LIMITATIONS: [Assumptions, power, generalizability]
Example
STATISTICAL ANALYSIS: Does treatment X reduce blood glucose vs. vehicle?
DATA CHARACTERISTICS:
Outcome: Continuous (mg/dL) | Groups: 2 (treatment vs. vehicle) | Distribution: Normal
Normality: Shapiro-Wilk W=0.97, p=0.34
Variance: Levene's F=0.12, p=0.73
TEST SELECTED: Independent t-test — normal distribution, equal variance, 2 independent groups
RESULTS:
t(46) = 3.82, p = 0.0004
Effect size: Cohen's d = 1.12 (large)
95% CI: [18.4, 59.2] mg/dL difference
n = 24 per group
INTERPRETATION: Treatment X significantly reduced fasting blood glucose by 38.8 mg/dL
(compared to vehicle), a large and clinically meaningful effect.
LIMITATIONS: Single timepoint, single strain, not adjusted for body weight.
1---2name: statistical-analysis-advisor3description: Advises on appropriate statistical tests, validates assumptions, performs analysis with effect sizes and confidence intervals. Outputs a statistical analysis plan or results interpretation with test justification and assumption checks.4allowed-tools: Read, Write, Bash5effort: medium6metadata:7 source: https://skillmd.com/skills/uitbreidenos/statistical-analysis-advisor8---910# Statistical Analysis Advisor1112## When to activate13When choosing statistical tests, interpreting results, checking assumptions, or when a reviewer questions your analysis. Use before running any analysis to ensure correct test selection.1415## When NOT to use16Skip for simple descriptive statistics (mean, SD), data visualization without inference, or when the statistical approach is already validated and approved.1718## Instructions19201. **Characterize the data:**21 - Outcome type: Continuous, categorical, ordinal, time-to-event, count22 - Groups: Number of groups, paired vs. independent23 - Distribution: Normal (Shapiro-Wilk), equal variance (Levene's)24 - Sample size per group25262. **Test selection decision tree:**27 - 2 independent groups, normal → Independent t-test28 - 2 independent groups, non-normal → Mann-Whitney U29 - 2 paired groups, normal → Paired t-test30 - 3+ groups → One-way ANOVA (Kruskal-Wallis if non-normal)31 - Repeated measures → RM-ANOVA or Friedman32 - Time-to-event → Log-rank test, Cox regression33 - Correlation → Pearson (normal) or Spearman (non-normal)34353. **Assumption checks:**36 - Normality: Shapiro-Wilk (n<50) or Kolmogorov-Smirnov (n≥50)37 - Homogeneity of variance: Levene's test38 - Sphericity: Mauchly's test (for RM-ANOVA)39 - If violated: Report the violation and use the non-parametric alternative40414. **Report results with:**42 - Test statistic and degrees of freedom43 - Exact p-value (not just <0.05)44 - Effect size (Cohen's d, η², r, OR as appropriate)45 - 95% confidence interval46 - Sample size47485. **Multiple comparisons:** If >2 comparisons, specify correction:49 - Bonferroni (conservative)50 - Holm-Bonferroni (step-down)51 - Benjamini-Hochberg FDR (exploratory)5253## Output Format5455```56STATISTICAL ANALYSIS: [Research question]5758DATA CHARACTERISTICS:59 Outcome: [type] | Groups: [X] | Distribution: [normal/non-normal]60 Normality: Shapiro-Wilk W=[X], p=[X]61 Variance: Levene's F=[X], p=[X]6263TEST SELECTED: [Test name] — [justification]6465RESULTS:66 [Test]([df]) = [statistic], p = [exact p]67 Effect size: [Cohen's d / η² / r] = [value] ([small/medium/large])68 95% CI: [[lower], [upper]]69 n = [X] per group7071INTERPRETATION: [Plain language conclusion]72LIMITATIONS: [Assumptions, power, generalizability]73```7475## Example7677```78STATISTICAL ANALYSIS: Does treatment X reduce blood glucose vs. vehicle?7980DATA CHARACTERISTICS:81 Outcome: Continuous (mg/dL) | Groups: 2 (treatment vs. vehicle) | Distribution: Normal82 Normality: Shapiro-Wilk W=0.97, p=0.3483 Variance: Levene's F=0.12, p=0.738485TEST SELECTED: Independent t-test — normal distribution, equal variance, 2 independent groups8687RESULTS:88 t(46) = 3.82, p = 0.000489 Effect size: Cohen's d = 1.12 (large)90 95% CI: [18.4, 59.2] mg/dL difference91 n = 24 per group9293INTERPRETATION: Treatment X significantly reduced fasting blood glucose by 38.8 mg/dL94(compared to vehicle), a large and clinically meaningful effect.95LIMITATIONS: Single timepoint, single strain, not adjusted for body weight.96```