RMP.5:4.2 - Use the analysis that fits the question
Apply the justified disciplinary Method. Its assumptions and criterion must come from the relevant practice, rather than from the prestige of a technique. The following contrasts help select the examination; they are not a list of procedures that every inquiry must perform.
| Research use | Practical analytical move | Criticism that can change the answer |
|---|---|---|
| Quantitative or experimental estimate | Recover the target quantity, sampling or assignment unit and comparison. Estimate the relevant magnitude with a justified uncertainty account. Preserve planned choices and consequential alternatives, exclusions and multiple comparisons. | Dependence, missingness, measurement, design assumptions or analytical choices can change magnitude, uncertainty or interpretation. A significance label alone states neither effect size nor causal warrant. |
| Observational or field account | Relate observations to their sites, periods, selection and measurement conditions. Separate the observed pattern from the proposed explanation. | Confounding, seasonality, selection or an absent comparison can prevent causal attribution even when the observed pattern is clear. Repeated readings of one site do not create independent sites. |
| Qualitative or reflexive interpretation | Use the selected approach and theoretical commitments to develop and ground the interpretation. Connect it to the material, the researcher’s contribution and relevant discordant or negative cases. | A theme can outrun its extracts, suppress a different account or imply unsupported transfer. Reflexive thematic analysis does not acquire methodological coherence from a coder-agreement score imported from another approach. |
| Historical or source-critical account | Examine dating, authenticity, provenance, source purpose, relationships among accounts and the sequence of the events claimed. | Copies can share one source; survival and access can explain a silence; a record may establish a date without establishing causation. Inspect contrary chronology and the limits of the archive actually examined. |
| Computational or model claim | Recover the represented target, derivation or implementation, data and environment, fitted quantities and relevant parameter or structural assumptions. Interpret the output for the stated target. | Correct execution can coexist with an inadequate model-to-world relation. Calibration, implementation verification, sensitivity and use-specific validation answer different questions. |
| Mixed inquiry | Analyze each component under its own Method and state the relation claimed between the results. | Dependence, different meanings, disagreement or non-comparable targets can limit the combined inference. Do not average away a disagreement that changes what the components jointly support. |
A descriptive estimate need not manufacture a causal design. A qualitative approach need not borrow statistical representativeness, universal saturation or reliability procedures that it does not justify. Historical criticism does not require recreating a unique past event. A missing domain Method or criterion limits the dependent analysis and returns to its competent supplier.