1 Result types in research workflows
1.1 Definition and purpose of result type
A result type is a structured label for the form in which outcomes are produced and interpreted in a study, experiment, or analysis pipeline. It specifies what the analysis yields—such as numerical summaries, category labels, ranked outputs, or qualitative themes—and, implicitly, what kinds of claims are appropriate for those outputs. In practice, the result type links analytic procedures to reporting conventions so that readers can understand both the evidence and the limits of inference.
1.2 When result type is specified (planning vs. analysis vs. reporting)
Result types are often defined at multiple points in a workflow. During planning, teams select primary and secondary outcomes, determine what form they will take (e.g., scores versus themes), and outline the computations required to produce them. During analysis, analysts may refine which result types are generated from the same underlying data, such as producing both inferential estimates and descriptive summaries. During reporting, authors finalize how those results are presented—tables, figures, narratives, or decision-oriented summaries—ensuring that interpretation remains consistent with the chosen type.
1.3 Relationship to measurement and data formats
Result types are constrained by measurement choices and by how data are stored. A study that measures a variable on a continuous scale naturally supports continuous numeric summaries and regression-based outputs, while categorical coding supports counts, proportions, and theme-based qualitative reporting. Similarly, data formats such as time series, cross-sectional tables, or coded text dictate which result types are convenient or valid. Although result type and data format are distinct concepts, they are interdependent: changing one often requires adjusting the other.
2 Common quantitative result types
2.1 Numeric summary results
2.1.1 Continuous outcomes (means, medians, trajectories)
Continuous result types express quantities that vary along a continuum. Common examples include measures of central tendency (mean or median), dispersion (variance or interquartile range), and temporal trajectories (how values change across time). Trajectories may be summarized with descriptive curves or with model-based time effects, depending on the analysis plan.
2.1.2 Discrete outcomes (counts, rates)
Discrete result types represent outcomes that occur in distinct categories or events, often summarized as counts or rates. Counts might include the number of occurrences in a period, while rates normalize by exposure such as time or population size. These outputs frequently serve as inputs to inferential procedures, including comparisons between groups or baseline-adjusted models.
2.2 Inferential results
2.2.1 Effect sizes and contrasts
Inferential results frequently include effect sizes—quantities intended to summarize the magnitude of differences or associations. Effect sizes and contrasts may be presented as differences in means, ratios, standardized measures, or model-derived comparisons. The purpose is to support substantive interpretation beyond mere statistical significance, offering a scale for how large an effect appears to be.
2.2.2 Uncertainty quantification (confidence intervals, credible intervals)
Uncertainty-aware result types express how precisely the analysis estimates its targets. Confidence intervals and credible intervals summarize uncertainty around parameters or effects, reflecting sampling variation or posterior uncertainty. Reporting such intervals helps readers gauge whether estimated values are stable and how sensitive conclusions might be to noise.
2.2.3 Hypothesis test outcomes (p-values, test statistics)
Some workflows include hypothesis-test result types, such as p-values and test statistics. These outputs address whether observed patterns are consistent with a null hypothesis under a specified model and error structure. While widely used, they are interpretively constrained by assumptions, model specification, and the choice of test procedure.
2.3 Model-based results
2.3.1 Predictions and forecasts
Model-based result types include predictions and forecasts, typically expressed as expected values for new cases or future time points. Predictions may be point estimates, interval forecasts, or full predictive distributions. They are used for tasks such as forecasting demand, estimating risk, or evaluating expected outcomes under a model.
2.3.2 Classification scores (probabilities, logits)
Classification-oriented result types transform model outputs into interpretable scores, such as probabilities or logits. Probabilities provide an intuitive scale for the likelihood of class membership, while logits are often used internally and later mapped to probabilities through a link function. The selected representation affects downstream decisions and calibration procedures.
2.3.3 Regression coefficients and partial effects
Regression result types may include coefficients that quantify the association between predictors and outcomes within a specified model. Partial effects extend this by translating coefficients into marginal or conditional impacts that may vary across covariate values. These outputs support interpretation of how predictors relate to the response, though the meaning depends on model assumptions.
2.4 Agreement and performance metrics
2.4.1 Accuracy, precision, recall
Performance result types evaluate predictive or classification systems relative to labeled data. Accuracy summarizes overall correctness, while precision and recall focus on different aspects of error: precision emphasizes correctness among predicted positives, and recall emphasizes coverage of true positives. These measures support comparative evaluation across models and datasets.
2.4.2 Calibration and ROC/AUC
Calibration-related result types assess whether predicted probabilities correspond to observed frequencies. Poor calibration may produce overconfident or underconfident scores even when discrimination is adequate. ROC curves and AUC provide discrimination-oriented summaries by measuring separability across thresholds.
2.4.3 Reliability and inter-rater agreement
When outcomes depend on human judgment, result types may report reliability. Inter-rater agreement metrics describe the consistency of coders or raters. These outputs are particularly relevant in annotation tasks and qualitative coding workflows, where reproducibility depends on how uniformly categories are applied.
3 Common qualitative result types
3.1 Thematic results
3.1.1 Codes and categories
Thematic result types are frequently built from coding schemes that map textual or observational material to codes and categories. Codes represent recurring features, while categories organize codes into higher-level groupings. The product of this process is an interpretive structure that supports later synthesis and comparison.
3.1.2 Themes and narrative summaries
Beyond coding, analyses may produce themes: higher-order claims about patterns of meaning within the data. Narrative summaries translate these themes into readable accounts, often combining evidence excerpts with interpretive statements. The resulting outputs aim to preserve context while summarizing recurring ideas.
3.2 Descriptive qualitative outputs
3.2.1 Field notes and memos
Descriptive qualitative result types include field notes and analytic memos. Field notes capture observations and context, while memos document reasoning, emerging hypotheses, and methodological reflections. Although often not treated as “final results” in the strictest sense, these outputs form an evidentiary trail that supports transparency.
3.2.2 Case descriptions and typologies
Case descriptions present structured accounts of individual units such as participants, sites, or events. Typologies further organize cases into conceptual types based on shared characteristics. These results are commonly used when the research goal is to characterize heterogeneity rather than estimate a single numeric parameter.
3.3 Interpretive or explanatory results
3.3.1 Mechanism narratives
Mechanism narratives describe plausible pathways or processes that connect causes to observed outcomes. Rather than only listing correlations, mechanism outputs attempt to explain how and why patterns occur, often grounded in empirical evidence from the dataset and linked reasoning.
3.3.2 Conceptual model outputs
Conceptual model result types express relationships among constructs, sometimes as diagrams, frameworks, or propositions. These outputs can summarize how variables or concepts interact, either as a hypothesis for future work or as an explanatory account derived from the current evidence.
4 Mixed-methods and integration-oriented result types
4.1 Joint displays and integrated summaries
4.1.1 Convergence and comparison matrices
Integration-oriented result types often use joint displays to organize quantitative and qualitative findings together. Convergence and comparison matrices align categories, themes, or variables across methods, highlighting agreement, partial overlap, or divergence. These displays support systematic interpretation rather than relying on informal juxtaposition.
4.1.2 Explanatory linking (connecting quantitative and qualitative strands)
Some integrated outputs emphasize explanatory linking, where qualitative insights are used to interpret quantitative patterns. For example, a theme may explain why a measured effect appears in one group but not another. This result type clarifies how the two strands inform each other at the level of mechanisms, context, or meaning.
4.2 Meta-integration outputs
4.2.1 Integrated themes with supporting metrics
Meta-integration result types combine thematic conclusions with quantitative support. A theme may be reported alongside associated frequency counts, effect estimates for related subgroups, or measures of strength and prevalence. This pairing is intended to make interpretive claims more concrete while preserving qualitative nuance.
4.2.2 Triangulated conclusions and boundary conditions
Triangulated conclusions aim to synthesize findings from multiple methods into claims with clearer conditions. Boundary conditions specify where a conclusion is likely to hold or fail, such as variations across contexts, subpopulations, or measurement regimes. This result type supports more cautious and better-scoped generalizations.
5 Decision and action-oriented result types
5.1 Decision rules and thresholds
5.1.1 Classification thresholds
Threshold-based result types convert scores into categorical actions. For instance, predicted probabilities may be mapped to “accept” or “reject” using a cutoff. The chosen threshold typically reflects trade-offs between false positives and false negatives and should be linked to the decision context.
5.1.2 Risk tiers and stratification outputs
Risk tiering stratifies subjects or instances into ordered groups, such as low, medium, and high risk. Stratification outputs can be based on absolute risk, relative ranking, or calibrated score intervals. These forms are used when actions differ by risk level and when stakeholders require interpretable groupings.
5.2 Intervention or recommendation outputs
5.2.1 Ranking of options
Recommendation result types may produce ranked options, such as which interventions appear most suitable given a model. Rankings can be driven by predicted outcomes, utility functions, or expected benefit under constraints. They are often accompanied by justification summaries tied to key features or outcomes.
5.2.2 Personalized outputs (when applicable)
Personalized result types tailor outputs to individual cases, using covariates to generate case-specific recommendations or expected outcomes. Such results require careful attention to validity, since performance can vary across subgroups and the meaning of personalization depends on the modeling framework and evaluation strategy.
5.3 Uncertainty-aware recommendations
5.3.1 Sensitivity statements
Sensitivity result types describe how recommendations change when assumptions, inputs, or model choices vary. Sensitivity statements help distinguish robust decisions from those highly dependent on uncertain elements, such as imputed variables, parameter settings, or calibration drift.
5.3.2 Scenario-based outputs
Scenario-based result types present recommendations under multiple plausible conditions. Rather than relying on a single forecast, analysts generate alternative futures or parameter regimes and report how guidance changes across scenarios. This format supports planning and risk management by showing what would need to change for conclusions to reverse.
6 Reporting result types effectively
6.1 Alignment between result type and research question
Effective reporting begins with matching result type to the research question. If the question asks for magnitude, effect estimates and uncertainty intervals are often appropriate. If the question asks for meaning or experience, thematic and interpretive outputs better fit. Misalignment can lead to outputs that are technically correct yet substantively uninformative.
6.2 Standard formatting conventions
6.2.1 Tables and figures by result type
Standard conventions vary by result type. Numeric summaries commonly appear in tables with measures of central tendency and dispersion, while uncertainty intervals are often depicted with error bars or interval plots. Classification performance is typically shown via ROC curves, precision-recall curves, or confusion matrices. Qualitative results are commonly presented with code summaries, theme tables, or structured narrative excerpts.
6.2.2 Labeling units, scales, and categories
Labels and metadata are essential for interpretability. Reports should specify units, scale direction, transformations, and the definition of categories. For numeric outputs, the meaning of higher versus lower values, the origin of scales, and the handling of missing data should be explicit. For categorical or coded qualitative results, definitions of codes and inclusion criteria for categories help readers understand how labels were assigned.
6.3 Documentation for reproducibility
6.3.1 Analysis artifacts and output schemas
Reproducibility is improved by documenting the artifacts that generate each result type. Output schemas—such as how variables map to tables or how model scores are stored—reduce ambiguity for future analysts. This includes reporting derived variables, preprocessing steps that affect result types, and the software components that produced them.
6.3.2 Versioning and provenance notes
Provenance involves recording where inputs, models, and parameters came from, and which versions were used. Versioning of datasets, code, and model configurations supports verification, especially when results depend on preprocessing choices or hyperparameters. Provenance notes help distinguish genuine analytical changes from artifacts of evolving pipelines.
6.4 Interpreting result types correctly
6.4.1 Common misinterpretations and how to avoid them
Misinterpretations often arise when readers treat result types as having the same meaning across contexts. Examples include confusing p-values with effect size, reading probabilities as calibrated without checking calibration, or treating themes as universally representative without documenting sampling and saturation logic. Clear definitions and careful wording can prevent these errors.
6.4.2 Communicating limitations tied to result type
Every result type carries typical limitations. Numeric summaries may hide subgroup variation; uncertainty intervals depend on model assumptions; predictive performance metrics may not reflect real deployment conditions; qualitative themes can be influenced by coder perspective and sampling. Reporting should connect limitations directly to the chosen result type, avoiding blanket statements not supported by the evidence.