statistical-design-audit

Auditing multiple comparisons requires checking denominator correctness and applying appropriate corrections. The audit validates your p-value formulas and bootstrap methods against the design's statistical assumptions, returning a PASS, CONCERN, or FAIL status with improvement insights.

Permutation inference requires strict null exchangeability and correct denominators to be valid. The audit checks these clustering and exchangeability assumptions to verify that your permutation design matches the statistical properties of the data.

Yes, statistical audits are designed for complex models involving multiple comparisons. The audit checks denominators, validates estimands, and verifies that bootstrap methods and multiple comparison corrections are appropriately applied.

P-value analysis fails on clustered data when the denominator formula ignores the data's clustering structure. The audit detects this mismatch by checking null exchangeability and Markov order, returning a FAIL status with insights to fix the denominators.

You should audit statistical design for data validity before trusting any inference results. The audit evaluates denominators, p-value formulas, and multiple comparison corrections to ensure the design matches statistical assumptions and prevents inference errors.

The best way to check estimand validity is to run a statistical design audit that evaluates denominators, p-value formulas, and multiple comparisons. This ensures the estimand aligns with the statistical assumptions and prevents inference errors before production.