Screen test assumptions before computing any inference

Role

data analyst

Task

Before analysing {{dataset_description}} to compare {{groups}} on {{outcome_variable}}, screen the assumptions of the proposed test {{proposed_test}}. Check measurement scale, distributional shape, variance homogeneity, and independence given the design, then either confirm the test or recommend a specific alternative and explain how the conclusion could change.

Context

The variable is measured on a {{scale_type}} scale and observations may be clustered by {{clustering_unit}}.

Inputs

• Dataset description and design: {{dataset_description}}
• Variable definitions and scale type {{scale_type}}
• Group sizes for {{groups}}
• Proposed test: {{proposed_test}}

Constraints

• Check each assumption separately with the diagnostic to be used
• Address independence explicitly given clustering by {{clustering_unit}}
• If recommending a nonparametric or model-based alternative, state what it tests instead
• Warn that conclusions can reverse when an inappropriate test is replaced

Output Format

An assumption-by-assumption checklist with diagnostics and verdicts, a test recommendation, and a sensitivity note on how conclusions could change.

Quality Criteria

• Each assumption paired with a concrete diagnostic
• Independence assessed against the actual design
• Alternative test's estimand described, not just named
• Possibility of conclusion reversal stated