How Do You Decide Which Characteristics Are Important Enough to Describe?

How Do You Know Which Characteristics Deserve a Place in Your Descriptive Study?

You are designing a study of university students' experiences with generative artificial intelligence. Your initial questionnaire includes age, sex, academic program, year level, device ownership, internet connectivity, frequency of AI use, preferred applications, academic performance, perceived usefulness, trust, anxiety, ethical concerns, and dozens of other characteristics.

Most of these variables seem potentially interesting. Some appear in previous studies. Others might be useful if an unexpected pattern emerges. Before long, the questionnaire has become considerably larger than the research question that motivated it.

Which characteristics are essential, which are merely useful, and which should be excluded?

This is not simply a matter of shortening the instrument. Selecting characteristics determines what the study can describe, how its findings can be interpreted, and whether the collected information justifies the time and burden imposed on participants.

Select Characteristics According to Their Contribution to the Research Question

How to Determine Which Characteristics Belong in a Descriptive Study

Begin With What the Study Is Supposed to Describe

The selection of characteristics should begin with the research purpose rather than the availability of questionnaire items or variables in an existing dataset.

A study describing students' access to digital learning resources requires different characteristics from a study describing their AI literacy, academic engagement, or experiences of online assessment.

Consider the question: "What barriers do undergraduate students encounter when accessing online learning materials?"

Relevant characteristics might include device availability, internet reliability, accessibility requirements, and reported difficulties accessing course materials. Students' preferred social media platforms might be interesting, but their relevance would require an additional justification.

The appropriate starting point is to identify the evidence that would actually answer the research question. Each proposed characteristic should then be evaluated against that requirement.

This approach reduces the risk of collecting numerous variables while failing to measure the phenomenon adequately.

Distinguish Characteristics, Variables, Constructs, and Indicators

These terms are related, but they are not interchangeable.

A characteristic is a property or attribute of a unit being studied. A variable is its measurable or classifiable representation, which may vary across units or observations. A construct is a concept that may not be directly observable, such as academic engagement or AI literacy. An indicator is an observable measure used to represent some aspect of that construct.

A construct may require several indicators. Conversely, one indicator may represent only a limited aspect of a broader construct.

For example, counting the number of AI applications a student has used does not necessarily measure AI literacy. Frequency of use, knowledge, critical evaluation, and responsible application are conceptually distinct, even when they are empirically related.

Researchers should therefore avoid selecting convenient indicators merely because they are easy to quantify.

Identify the Core Characteristics Before Adding Contextual Variables

Core characteristics are those without which the principal research question cannot be answered.

If the objective is to describe the frequency and purposes of generative AI use among students, the study must measure usage frequency and relevant purposes of use. These characteristics form the substantive center of the investigation.

Contextual variables, such as academic program or year level, may help describe how the findings vary across population categories. They can be valuable, but they should not automatically receive the same priority as the primary outcomes.

Methodological variables may also be necessary. Sampling strata, survey mode, or eligibility information can support population estimation even when they are not central findings.

This distinction helps researchers organize the study without assuming that every collected variable must become a separate research objective.

Use a Conceptual Framework Without Letting It Dictate Every Variable

Conceptual frameworks can help identify relevant dimensions of a phenomenon and clarify how characteristics are understood.

For example, a study of teachers' AI literacy might draw on a published framework that distinguishes knowledge, practical competencies, and critical evaluation. Such a framework can help identify dimensions that a narrow measure of tool familiarity would overlook.

However, adopting a framework does not necessarily require measuring every concept it contains. Some dimensions may fall outside the study's objectives, while others may not be measurable adequately with the available resources.

Researchers should explain which dimensions are included, why they matter, and what the selected measures can reasonably represent.

They should also remain open to characteristics that are important in the study context but insufficiently represented in an existing framework.

The framework should improve conceptual coverage rather than function as an unquestioned checklist.

How Do You Judge Whether a Characteristic Is Substantively Important?

Substantive importance concerns the contribution a characteristic makes to understanding the phenomenon or informing a relevant decision.

A characteristic may deserve inclusion because it captures a central dimension of the research question, identifies a meaningful population difference, or provides information necessary for educational planning or policy.

For instance, internet reliability may be more consequential than preferred device brand in a study intended to inform digital learning access policies.

Importance should not be equated with statistical variability. A characteristic that varies little may still reveal a consequential population condition. If nearly every student lacks a particular accessibility accommodation, that finding could be important even though the measured characteristic shows limited variation.

Similarly, characteristics relevant to a small or disadvantaged subgroup may warrant attention despite their limited contribution to an overall population average.

Researchers should therefore consider both the central research purpose and the potential consequences of overlooking important experiences or population segments.

Does Every Characteristic Need Support From Previous Literature?

Previous research provides an important basis for identifying constructs, selecting indicators, and evaluating measurement approaches. It can also reveal which characteristics have been consistently associated with the phenomenon or used in comparable descriptive studies.

However, prior use alone is not sufficient justification for inclusion. A variable that was important in another population or research context may have limited relevance to the present study.

Conversely, emerging phenomena may involve characteristics that have not yet been extensively examined. Researchers can justify such characteristics through conceptual reasoning, preliminary qualitative evidence, stakeholder consultation, or the practical requirements of the research question.

The absence of an extensive literature does not automatically make a characteristic irrelevant. It does increase the importance of explaining why the characteristic is being measured and how it will be operationalized.

When using newly developed measures, researchers should distinguish exploratory measurement from established instrument validity.

Evaluate Whether Each Characteristic Can Be Measured Adequately

A concept may be important but difficult to measure reliably within the available design.

Suppose researchers want to describe students' responsible use of generative AI. Asking whether respondents consider themselves responsible users may provide information about self-perception, but it does not necessarily establish responsible behavior.

More defensible measurement might require clearly defined practices, scenario-based judgments, or multiple indicators. The appropriate choice depends on how the construct is conceptualized.

Measurement quality should be considered in terms of validity, reliability where applicable, interpretability, and suitability for the target population.

Validity concerns the evidence supporting the intended interpretation and use of scores or observations. Reliability concerns consistency or precision under specified measurement conditions. A measure can be consistent without adequately representing the intended construct.

Researchers should also consider whether respondents can reasonably provide the requested information. Questions about events occurring several years earlier may be vulnerable to recall error, while sensitive behaviors may be affected by social desirability.

How Many Characteristics Should a Descriptive Study Include?

There is no universal number of variables that makes a descriptive study sufficiently comprehensive or excessively broad.

A study describing a narrowly defined outcome may require only a small set of measures. A national population survey intended to support several policy decisions may legitimately require a much broader instrument.

The relevant issue is whether each characteristic has a clear function and whether the combined measurement burden remains proportionate to the study's purpose.

Adding variables can increase questionnaire length, respondent fatigue, missing data, analysis complexity, and the possibility of producing numerous findings without a coherent interpretive framework.

At the same time, removing important contextual characteristics may prevent meaningful subgroup descriptions or obscure inequalities.

The objective is not to minimize the number of variables mechanically, but to avoid collecting so many unrelated characteristics that the descriptive study loses focus.

Should Demographic Variables Always Be Included?

No demographic variable is universally required in every descriptive study.

Age, sex, educational level, institutional affiliation, socioeconomic indicators, and other characteristics may be relevant when they define the target population, support sampling adjustments, or allow substantively meaningful descriptions.

However, collecting demographic information without a clear purpose can create unnecessary privacy risks and analytical obligations.

For example, a study describing the technical performance of a learning management system may not need detailed personal demographic information if the unit of analysis is the system transaction rather than the student.

When demographic information is necessary, researchers should consider inclusive response options, appropriate category definitions, data protection, and the possibility of identifying individuals through combinations of characteristics.

Particular care is warranted when subgroup sizes are small or characteristics are sensitive.

Which Characteristics Are Needed for Meaningful Subgroup Analysis?

Subgroup variables should be selected when differences across population categories are relevant to the research objectives or necessary for interpreting the findings.

For example, a university-wide estimate of digital access may conceal substantial differences between students living in university housing and those commuting from distant areas.

Describing these categories separately may reveal conditions that are not apparent in the overall average.

However, subgroup selection should not become an unrestricted search for interesting differences. Researchers should identify important subgroup comparisons in advance when possible, assess whether sample sizes support interpretable estimates, and report uncertainty appropriately.

Small subgroups may require deliberate oversampling or alternative reporting approaches. Privacy and disclosure risks may also limit the level of detail that can be published.

Descriptive subgroup comparisons should not be interpreted as causal effects merely because differences are observed. The distinction between describing group differences and explaining their causes remains important.

Can Existing Datasets Determine Which Characteristics You Study?

Secondary datasets can provide valuable descriptive evidence, particularly when they offer broad population coverage or repeated measurements.

Nevertheless, researchers should not assume that available variables are adequate representations of the intended constructs.

An institutional dataset may record whether students accessed the learning management system, but access logs do not necessarily measure engagement, comprehension, or learning quality.

When using existing data, researchers should inspect variable definitions, collection procedures, missingness, coverage, and changes in measurement over time.

If the dataset cannot measure an essential characteristic adequately, the appropriate response may be to narrow the research question, obtain supplementary evidence, or acknowledge the limitation.

Renaming an available proxy as though it directly measured a broader construct does not resolve the measurement problem.

What Role Should Stakeholders Play in Selecting Characteristics?

Stakeholder consultation can help researchers identify characteristics that matter in practice but may be overlooked by existing literature or institutional records.

In educational research, relevant perspectives might include students, teachers, academic administrators, disability support personnel, or others affected by the phenomenon.

For example, administrators may prioritize the proportion of students with internet access, while students may identify connection stability during timed examinations as a more consequential characteristic.

Consultation can therefore improve the relevance and content coverage of descriptive measures.

However, stakeholder preferences should not replace methodological judgment. Researchers must still evaluate whether proposed characteristics are conceptually appropriate, measurable, ethical, and aligned with the research objectives.

How Do You Balance Information Value Against Respondent Burden?

Every additional measure has potential benefits and costs. Its value depends on the information it contributes relative to the burden of obtaining it.

Longer questionnaires may increase completion time and fatigue. Complex or sensitive questions may also affect participation and response quality.

Researchers should consider whether information is already available from reliable administrative sources, whether several questions are unnecessarily repetitive, and whether the proposed measure will actually be used in analysis.

For some constructs, multiple items are necessary for adequate measurement. Removing them solely to shorten a questionnaire may compromise validity.

The appropriate balance is therefore construct-specific. A short but inadequate instrument is not necessarily preferable to a longer, well-justified one.

How Should Researchers Handle Characteristics That Emerge During the Study?

Descriptive research may reveal unexpected categories, experiences, or patterns that warrant additional attention.

For example, open-ended responses about barriers to online learning might identify shared-device arrangements as an important characteristic that the original questionnaire did not adequately capture.

Researchers may report this finding as exploratory and consider incorporating a more systematic measure in a subsequent study.

However, if a characteristic was not measured consistently across all participants, researchers should avoid presenting its observed frequency as a reliable population prevalence estimate.

When new variables or analytical categories are developed after inspecting the data, their exploratory status should be disclosed.

Transparent documentation helps readers distinguish prespecified descriptions from findings generated through the research process.

Prepare a Variable Justification Matrix Before Data Collection

A practical way to evaluate proposed characteristics is to document why each variable is needed, how it will be measured, and how its results will be used.

This does not require a complicated scoring system. A concise matrix can make the reasoning visible and reveal variables that lack a clear purpose.

Characteristics that satisfy several criteria may deserve high priority. Those that cannot be linked to any meaningful objective or methodological requirement should be reconsidered.

The matrix is a decision aid, not a validated numerical instrument. Researchers should not assign arbitrary scores and treat the resulting ranking as an objective measure of scientific importance.

Selecting Characteristics for a Study of Students' Digital Learning Access

The resulting questionnaire may contain fewer variables, but its descriptive coverage is more closely aligned with the research question.

The study is not stronger merely because it is shorter. It is stronger to the extent that the retained measures adequately represent the phenomenon, support valid interpretation, and avoid unnecessary collection of information.

Common Mistakes When Selecting Descriptive Characteristics

How to Decide Which Characteristics to Keep, Revise, or Remove

Before finalizing your instrument, list every proposed characteristic and connect it to a specific research question, population requirement, or analytical purpose.

Characteristics that directly answer the principal questions should ordinarily receive priority. Contextual variables should be retained when they support meaningful interpretation, population estimation, or justified subgroup descriptions.

Where a characteristic is important but poorly measured, revise the operational definition or measurement approach rather than retaining an inadequate indicator merely for convenience.

A useful final exercise is to imagine the results section before collecting the data. For each proposed characteristic, ask what descriptive estimate or finding would be reported and how it would contribute to answering the study's question.

If no defensible answer emerges, the variable may not deserve inclusion. Conversely, if a research objective cannot be answered with the proposed measures, the instrument remains incomplete regardless of how many variables it contains.

The goal is a descriptive design capable of producing meaningful and defensible evidence, not a questionnaire that measures everything researchers might conceivably find interesting.

Checklist for Selecting Characteristics in Descriptive Research

Frequently Asked Questions About Selecting Descriptive Characteristics

Measure What the Study Needs to Explain Clearly, Not Everything Available

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.