Practical guides · Page 18

Check Expected Counts Before a Contingency-Table Test

Audit category counts, independence, and sparse cells before interpreting a categorical association test.

Match each dataset column to a clear definition, value labels, missing-value rules, and a traceable derivation.

Build a labeled agreement table and interpret Cohen’s kappa alongside observed agreement and category frequencies.

Preserve ordered response labels, denominators, and scoring rules when describing ordinal research outcomes.

Preserve matched binary outcomes and separate directional change from overall agreement before testing.

Record generation settings and independent checks while preserving allocation concealment during enrollment.

Record SHA-256 digests to detect byte-level file changes while keeping content validation and access control separate.

Inspect unequal weights without confusing a simple effective sample calculation with full survey-design precision.

Find duplicate, ambiguous, and incompatible column names before renaming research data fields.

Plan paired differences, limits of agreement, and acceptable error before comparing two measurement methods.

Write explicit membership and grouping rules so participant counts remain traceable across efficacy, safety, and sensitivity analyses.

Compare statistical support by study fit, written scope, reproducible deliverables, data handling, and revision terms—with a downloadable checklist.