8. Quantitative Research: Measurement, Patterns, and Relationships
8.1 Purpose and design
Quantitative research uses numerical data to describe distributions, compare groups, estimate relationships, or test specified hypotheses. Designs may be descriptive, correlational, experimental, quasi-experimental, longitudinal, or cross-sectional. Design choice depends on the question and what inferences are intended. A cross-sectional survey measures variables at one point or period and is efficient for describing patterns, but temporal ordering and causal inference are often limited.
8.2 Concepts, constructs, and operationalisation
A construct such as engagement, trust, psychological safety, or project success is abstract and requires operationalisation: a reasoned plan for how it will be measured. Researchers should define the construct, justify the measure, consider its dimensions, and explain scoring. A single question may be inadequate for a complex construct. If an established scale is adapted, the adaptation and any licensing or permission requirements should be checked. Translation requires attention to meaning, not only literal wording.
8.3 Population, sampling frame, and sample
The population is the group to which the research question refers. A sampling frame is the accessible list or mechanism from which units are selected. Probability sampling gives units known selection probabilities and can support statistical generalisation under appropriate conditions. Non-probability methods, such as convenience or purposive sampling, may be practical but introduce different limitations. A large sample does not automatically eliminate selection bias or make a sample representative.
8.4 Sample-size reasoning
Sample size should be justified in relation to the research design, expected effect, variability, desired precision, number of predictors, missing data, and feasible recruitment. Power analysis may be appropriate for hypothesis testing; precision-based calculations may suit estimation. The researcher should state assumptions rather than present a number as universally correct. Response rate, clustering, design effects, and the quality of the sampling frame can matter as much as the nominal sample size.
8.5 Reliability and validity
Reliability concerns consistency of measurement; validity concerns whether evidence supports the intended interpretation and use of scores. Internal consistency statistics such as Cronbach’s alpha may be informative for suitable multi-item scales but do not prove that a scale is unidimensional or valid. Content, construct, criterion, and consequential considerations may all matter. Pilot testing can identify confusing wording, technical problems, response burden, and preliminary measurement concerns.
8.6 Descriptive and inferential analysis
Descriptive statistics summarise data through counts, percentages, means, medians, ranges, and dispersion. Inferential methods estimate uncertainty when drawing conclusions beyond observed data. Correlation measures association, not proof of causation. Regression estimates relationships conditional on included variables and model assumptions; omitted confounders, measurement error, multicollinearity, non-linearity, and selection effects can undermine interpretation. Researchers should report effect sizes and uncertainty, not only whether a p-value crosses a threshold.
8.7 Missing data and responsible interpretation
Missing data can arise from skipped items, attrition, system failures, or questions participants find sensitive. The pattern and likely causes should be examined; the chosen treatment must be justified. Deleting cases may reduce power or bias estimates if missingness is systematic. Researchers must avoid selectively reporting only favourable results, repeatedly testing until a significant result appears, or presenting exploratory findings as if they were pre-planned confirmatory tests. Transparent reporting is essential for credibility.