Library / Research Method Practice Principles Framework
Jump to passage
In this reading

Link to current text

Published source confirmed at last check

Source changed 2026-10-03 05:29:54 UTC · snapshot created 2026-10-03 05:30:57 UTC · last check 2026-10-03 06:05:20 UTC

RMP.2:8 - Common Anti-Patterns and How to Avoid Them

Anti-patternWhy it failsRepair
“Use the gold-standard design”Strength is question-, threat-, and feasibility-relative; one family cannot expose every error.Compare serious options by the exact criticism demand.
“The protocol is the design”A template can serialize choices without justifying them.Record the question-to-evidence relation, alternatives, threats, and selection rule first.
“Evidence is incomplete, therefore study”An unresolved claim does not select a useful obtainable acquisition.Compare the whole burden and contribution; return QualifiedNoNewStudy when the current allocation closes without a new study.
“No budget, therefore evidence sufficient”Resource refusal is not an epistemic conclusion.Preserve the bounded answer or unresolved question in QualifiedNoNewStudy when no current study is selected; resource and receiving decisions remain with their owners.
“Select provisionally and validate later”A missing mechanism, measure, access, safety, or authority result may overturn the family.Return ResearchDesignHold with the exact owner result and forked consequences.
“Everything must be fixed before selection”Some implementation details cannot change the selected relation.Prove invariance across the bounded admissible values and record PostSelectionNeed.
“Two software packages are two designs”Tool variation may leave evidence and criticism unchanged.Require an answer-changing relation, threat, or refusal difference.
“Mixed means stronger”Components can disagree, duplicate bias, or lack a warranted integration relation.Name each component’s irreducible role and the rule for disagreement or failure.
“A rerun is a replication”Code/data/environment reproduction, new-data replication, validation, and robustness test different claims.Name the live threat and select the credibility move that bears on it.
“The selected design authorizes the study”Design justification does not grant access, safety, ethics, rights, resources, or receiving authority.Keep every external owner result and later decision separate.