BdM-15/capture-insights / skills/data-analyzer
data-analyzer
VerifiedAdvanced data analysis, pattern detection, and insight generation from structured and unstructured datasets. Use when the user wants to analyze data, perform statistical analysis, find insights, detect patterns, identify anomalies, compare segments, test hypotheses, or generate data-driven recommendations. Triggers on phrases like 'analyze data', 'data analysis', 'find insights', 'analyze dataset', 'statistical analysis', 'find patterns', 'compare groups', 'test hypothesis', 'correlation analysis', or 'trend analysis'.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 8.6 KB |
| assets/report_template.md | markdown | 1.7 KB |
| evals/evals.json | config | 2.0 KB |
| references/pitfalls.md | markdown | 4.0 KB |
| references/statistical-methods.md | markdown | 4.8 KB |
| references/visualization-guide.md | markdown | 3.6 KB |
references/pitfalls.md
View on GitHub# Common Analytical Pitfalls
## Statistical Pitfalls
### P-hacking
**What it is:** Testing multiple hypotheses and only reporting the significant ones.
**Why it matters:** With α = 0.05, 1 in 20 tests will appear significant by chance. Running 20 tests and reporting the one that "worked" is not valid analysis.
**Fix:** Pre-register hypotheses before looking at data. Apply Bonferroni correction or FDR control when running multiple tests.
### Cherry-picking Data
**What it is:** Selecting the subset of data that supports a predetermined conclusion.
**Fix:** Define your data selection criteria before analysis. Document every exclusion. Test sensitivity — does the finding survive if you include the excluded data?
### Ignoring Assumptions
**What it is:** Using parametric tests (t-test, ANOVA, Pearson) on data that violates their assumptions.
**Fix:** Always check normality (Shapiro-Wilk) and homogeneity of variance (Levene's) before applying parametric tests. Use non-parametric alternatives if violated.
### Misinterpreting P-values
**Common misreading:** "p < 0.05 means there is a 95% chance the hypothesis is true."
**Correct interpretation:** If H0 were true, there is a less than 5% probability of observing results this extreme or more extreme by chance.
**Fix:** Always pair p-value with effect size and confidence interval.
### Focusing Only on Statistical Significance
**What it is:** Reporting "significant" results from large samples where tiny, meaningless differences reach p < 0.05.
**Fix:** Report effect size (Cohen's d, R², η²) alongside p-value. A result can be statistically significant but practically irrelevant.
---
## Logical Pitfalls
### Confusing Correlation and Causation
**What it is:** Assuming A causes B because they are correlated.
**Classic example:** Ice cream sales correlate with drowning rates — both are caused by hot weather.
**Fix:** Identify the mechanism. Look for confounders. Consider natural experiments or controlled comparisons.
### Survivorship Bias
**What it is:** Analyzing only the cases that "made it" and ignoring failures.
**Example:** Studying successful companies to learn what makes companies succeed — failed companies aren't in the sample.
**Fix:** Explicitly ask "what data is missing from this dataset?" before drawing conclusions.
### Simpson's Paradox
**What it is:** A trend that appears in subgroups disappears or reverses when groups are combined.
**Example:** A treatment appears effective in both men and women separately, but ineffective in the combined group (due to different group sizes).
**Fix:** Always segment data before combining. Check whether aggregate conclusions hold within subgroups.
### Base Rate Neglect
**What it is:** Ignoring the baseline frequency when interpreting rates or probabilities.
**Fix:** Always contextualize rates with the base rate population. Ask "what would we expect by chance?"
---
## Data Handling Pitfalls
### Ignoring Missing Data
**What it is:** Assuming data is Missing Completely at Random (MCAR) when it may be Missing at Random (MAR) or Missing Not at Random (MNAR).
**Fix:** Analyze the missingness pattern. Is data more likely to be missing for certain groups? If MNAR, simple imputation will bias results.
### Overfitting
**What it is:** Building a model complex enough to fit the training data perfectly but that fails to generalize.
**Fix:** Use cross-validation. Apply regularization. Prefer simpler models when performance is similar.
### Extrapolating Beyond the Data Range
**What it is:** Making predictions outside the range of observed values.
**Fix:** Clearly state the range of the data. Mark predictions outside that range as extrapolations with explicit uncertainty.
### Data Snooping
**What it is:** Examining the data before deciding on the analysis approach, then choosing the approach that gives the most interesting result.
**Fix:** Separate exploratory analysis (hypothesis generation) from confirmatory analysis (hypothesis testing). Use a holdout set for confirmation.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |