Assumption Light Inference¶
Use inference methods that require fewer fragile assumptions when strong assumptions are unjustified.
The Diagnostic Story¶
Symptom: The analysis uses a familiar method and returns a precise-looking number, but the data are small and skewed, or ordinal, or outlier-heavy, and the method assumes none of that. Different analysts making equally defensible choices get different answers. Reviewers argue about model validity without pinning down which assumption is actually doing the work.
Pivot: Turn assumption dependence into a visible design choice. Inventory the assumptions behind the planned inference, identify which ones are consequential and least justified, choose a procedure that requires fewer fragile commitments, and frame the result only within the limits that can actually be defended.
Resolution: Conclusions become less brittle because they no longer hang on unverified structural assumptions. Where a robust method yields a weaker claim, that weakness is honest rather than hidden. Method disagreements become productive because they can be mapped to assumption burden rather than professional preference.
Reach for this when you hear…¶
[clinical trial stats] “The t-test says significant, but with twelve patients and two obvious outliers I want to see the Wilcoxon before I believe it.”
[survey research] “We treated a five-point satisfaction scale like interval data and got a tidy regression — nobody checked whether the distances between points actually mean anything.”
[financial risk analysis] “Our VaR model looked great until we hit a fat-tailed week and realized we'd been assuming normality the whole time.”
When This Archetype Applies¶
Partial catalog groundingSome structural conditions are represented by existing abstractions, but no sufficient condition set is fully represented.
Diagnostic problem
A decision, claim, or comparison is being inferred from evidence through assumptions that may not hold. The procedure may look precise because it uses a formal model, but the conclusion is fragile if the data are small, skewed, outlier-heavy, nonstationary, ordinal, dependent, censored, or otherwise mismatched to the assumed form.
What this problem means
A conclusion depends on more structure than the evidence can responsibly carry. A parametric model, ordinary mean comparison, default test, or standard interval may assume a distribution, independence pattern, measurement scale, or stable process that is not actually established.
The structural danger is false security. The result can look precise because the method is formal, while the inference is actually fragile because a hidden assumption is carrying the conclusion. The system then mistakes methodological familiarity for evidential strength.
Show the applicability expression
Applicability expression6 distinct conditions
groundedpartly groundedopen
Equivalent to the 5 condition sets it replaces, with 3 duplicate condition cards removed.
5At least one of theselettered A–B
Any one of these groups completes the pattern; conditions inside a group are required together.
Uncertain data process · grounded
The data-generating process is poorly understood, changing, or too heterogeneous for a strong parametric model to be trusted.
A parametric model, ordinary mean comparison, default test, or standard interval may assume a distribution, independence pattern, measurement scale, or stable process that is not actually established. The narrower requirement in this condition set is: The data-generating process is poorly understood, changing, or too heterogeneous for a strong parametric model to be trusted.
Assumption-dependent conclusion · open
The formal conclusion depends materially on the poorly supported or mismatched assumption identified by the condition.
This condition preserves a load-bearing part of the diagnostic problem that was not captured by a source-condition atom. It remains explicit because omitting it would weaken the sufficient condition set.
Nonstandard sample structure · grounded
The sample is small, skewed, heavy-tailed, ordinal, censored, clustered, dependent, or visibly affected by outliers.
The procedure may look precise because it uses a formal model, but the conclusion is fragile if the data are small, skewed, outlier-heavy, nonstationary, ordinal, dependent, censored, or otherwise mismatched to the assumed form. The narrower requirement in this condition set is: The sample is small, skewed, heavy-tailed, ordinal, censored, clustered, dependent, or visibly affected by outliers.
Assumption-dependent conclusion · also required in this branch
Same condition as B above — stated once.
Assumption-sensitive result · open
A formal result changes substantially when distributional assumptions, independence assumptions, or outlier handling are varied.
The result can look precise because the method is formal, while the inference is actually fragile because a hidden assumption is carrying the conclusion. The narrower requirement in this condition set is: A formal result changes substantially when distributional assumptions, independence assumptions, or outlier handling are varied.
Familiarity-driven method · open
A default method is being applied because it is familiar rather than because its assumptions match the evidence.
In that situation, the productive move is to surface the assumptions and choose a method or representation that asks less of the evidence. The narrower requirement in this condition set is: A default method is being applied because it is familiar rather than because its assumptions match the evidence.
Assumption-dependent conclusion · also required in this branch
Same condition as B above — stated once.
Overconfident confirmation · open
Exploratory work has found a pattern, but confirmation under a strong model would create false confidence.
A parametric model, ordinary mean comparison, default test, or standard interval may assume a distribution, independence pattern, measurement scale, or stable process that is not actually established. The narrower requirement in this condition set is: Exploratory work has found a pattern, but confirmation under a strong model would create false confidence.
Assumption-dependent conclusion · also required in this branch
Same condition as B above — stated once.
Other requirements and context (2)
Why these sit outside the expression
Goal — a goal states an intended outcome or evaluation criterion, not a pre-existing situation that independently summons the archetype.
Application gate — it governs whether applying the archetype is appropriate or material, rather than defining the structural problem itself.
GoalDecision-makers need an inference, comparison, or uncertainty statement but cannot justify the usual model requirements.
Use this archetype when a claim, comparison, or uncertainty statement must be made from evidence whose usual assumptions are doubtful. In this archetype, the relevant goal is: Decision-makers need an inference, comparison, or uncertainty statement but cannot justify the usual model requirements. It supplies a criterion for evaluating what the intervention should accomplish or preserve.
Application gateThe domain has high consequences for overclaiming, and a modest but robust conclusion is preferable to a precise but brittle one.
The result can look precise because the method is formal, while the inference is actually fragile because a hidden assumption is carrying the conclusion. In this archetype, the relevant application gate is: The domain has high consequences for overclaiming, and a modest but robust conclusion is preferable to a precise but brittle one. It narrows when choosing or applying the archetype is warranted or decision-relevant.
Coverage
2 of 6 conditions grounded · 4 open.
None of the 4 open conditions sit in the shared core — each falls inside one alternative branch, so grounding any one of them closes only that branch.
Mechanisms / Implementations¶
- Assumption Audit Checklist: Enumerates the assumptions a planned inference rests on and flags which ones would change the conclusion if they failed — before any test is run.
- Bootstrap-Like Checks: Resamples the observed data with replacement to see whether an estimate holds still — gauging stability without trusting a parametric error formula.
- Diagnostic Plot Review: Reads fitted-data graphics to see whether a method's distributional and scale assumptions actually hold, catching violations a summary statistic hides.
- Median-Based Summaries: Reports the middle and the spread with order statistics — median, quantiles, IQR — so a few extreme values can't dominate the typical-case claim.
- Model Comparison Table: Lays the same question's answers side by side under strong and assumption-light frames, turning method disagreement into a visible, decidable finding.
- Nonparametric Tests: Compares groups or distributions with distribution-free tests chosen against a named assumption threat, not by software default.
- Permutation Tests: Builds an exact null by reshuffling the labels the hypothesis says are exchangeable, replacing a distributional assumption with a randomization one.
- Rank-Based Methods: Replaces raw values with their order positions so an inference leans on defensible ranking rather than unverified metric distance.
- Robust Statistics: Estimates with outlier-resistant methods whose conclusions survive a handful of extreme observations, then reports what that resistance costs.
- Sensitivity Analysis Protocol
Related Abstractions¶
Abstractions this archetype builds on — directly (a source ingredient) or as a related pattern. Links follow the typed catalog namespace.
Built directly on (3)
- Nonparametric Methods: Distribution-free analysis.
- Robustness: Maintain functionality under stress.
- Uncertainty: Incomplete knowledge.
Also references 12 related abstractions
- Approximation: Good-enough representation.
- Boundedness: Values remain within limits.
- Confidence Intervals: Range of plausible values.
- Hypothesis Testing (Null vs. Alternative): Null vs alternative evaluation.
- Overfitting: Poor generalization.
- Parsimony (Occam's Razor): Prefer simplicity.
- Probability: Quantifies uncertainty and likelihoods.
- Reproducibility & Replicability: Repeatable results.
- Sensitivity Analysis (in Operations Research): Analyze impact of parameter variation.
- Stationarity: Stable statistical properties.
Variants¶
Narrower or domain-specific specializations that share this archetype's core structure. Recognized variants are established; candidate variants are provisional.
Rank-Based Inference · mechanism family variant · recognized
Uses order information rather than metric distance when ranks are more defensible than interval-scale assumptions.
Permutation-Based Inference · mechanism family variant · candidate
Uses label rearrangement or randomization logic to build a comparison reference without a parametric sampling distribution.
Outlier-Robust Inference · risk or failure variant · recognized
Uses summaries, comparisons, or estimators that reduce undue dependence on extreme observations while still respecting meaningful extremes.
Editorial Notes¶
Problem Classification¶
Classification: Uncertainty, Evidence & Inference Failure → Probability, Distribution & Risk Calibration
Problem kernel: formal inference depends on untested distributional assumptions
Rationale: Small, skewed, dependent, censored, outlier-heavy, or nonstationary data make a precise procedure fragile to its assumed statistical form
Independent corroboration: The earliest necessary condition in the frozen evidence is: A decision, claim, or comparison is being inferred from evidence through assumptions that may not hold. That is a probability distribution and risk calibration problem because Probability, uncertainty intervals, tails, multiplicity, and variability are interpreted under hidden frames or assumptions that misstate risk.
Review outcome: Independent reviewer agreement; medium confidence.