Unit 2: Research Process and Research Design

MGN206 — Research Methodology 10 min read

I. Foundations of Systematic Research

Research methodology is the systematic study of how research is planned, conducted, analysed, and evaluated. It connects a research problem with defensible evidence by specifying the procedures used to collect data, test explanations, control error, and draw conclusions.

Systematic inquiry: Research follows an ordered sequence, from identifying a problem to reporting findings, rather than relying on intuition or unstructured observation.
Empirical basis: Conclusions are grounded in observable or measurable evidence, such as survey responses, experimental outcomes, interviews, documents, or field observations.
Logical reasoning:
- Deduction: Begins with theory and derives a testable prediction; for example, a theory may predict that increased study time improves examination scores.
- Induction: Begins with observations and develops patterns, concepts, or theory from them.
Objectivity: Procedures should reduce the influence of the researcher's preferences through standardised instruments, explicit criteria, and transparent analysis.
Validity: A study must measure or explain what it claims to address. For example, a mathematics test should measure mathematical ability rather than reading difficulty.
Reliability: A measurement procedure should produce reasonably consistent results under comparable conditions.
Replicability: Methods should be documented sufficiently for another researcher to repeat or closely reproduce the investigation.
Ethical responsibility: Research involving people requires informed consent, privacy protection, minimisation of harm, and honest reporting.
Feasibility: The question and design must fit available time, funding, skills, participants, data access, and institutional requirements.
Generalisation: Researchers assess whether findings can extend beyond the observed cases to a larger population, setting, or theoretical class.

II. The Research Process — From Problem to Report

A. Research Process Steps

The research process is an interconnected sequence in which each step shapes the quality and interpretation of later decisions.

Identify and define the research problem: State the issue precisely by identifying the population, relevant variables, context, and knowledge gap.
- A broad concern such as “employee stress” may become: “What is the relationship between weekly working hours and perceived stress among hospital nurses?”
- A good problem is clear, researchable, significant, ethical, and feasible.
Review the literature: Examine existing theories, findings, methods, and disagreements connected with the problem.
- Purpose: The review prevents unnecessary duplication, reveals gaps, defines key concepts, and provides a basis for hypotheses.
- Concrete sources: Peer-reviewed articles, scholarly books, official reports, theses, and credible datasets are commonly examined.
- Synthesis: Sources are compared by themes and evidence rather than merely listed one after another.
Develop objectives and research questions: Convert the problem into statements that direct data collection and analysis.
- Objective: “To determine whether working hours predict stress scores among hospital nurses.”
- Research question: “How strongly are weekly working hours associated with perceived stress?”
- Objectives normally begin with operational verbs such as describe, compare, examine, or evaluate.
Formulate hypotheses where appropriate: A hypothesis is a testable prediction about variables.
- Null hypothesis ((H_0)): Proposes no difference or association.
- Alternative hypothesis ((H_1)): Proposes a difference or association.

TEXT

H₀: ρ = 0
H₁: ρ ≠ 0

Here, (H_0) is the null hypothesis, (H_1) is the alternative hypothesis, and (\rho) is the population correlation between working hours and stress.

Define and operationalise variables: Translate abstract concepts into measurable indicators.
- Independent variable: The presumed cause or predictor, such as weekly working hours.
- Dependent variable: The measured outcome, such as a score on a perceived-stress scale.
- Control variable: A factor such as age or job role that is held constant statistically or procedurally.
- Operational definition: “Working hours” may mean the total hours recorded on duty rosters during the previous seven days.
Select the research design: Choose the framework that best answers the question while controlling error.
- A causal question may require an experiment, whereas a question about present attitudes may require a cross-sectional survey.
- The design specifies timing, comparison groups, measurement procedures, and the degree of researcher intervention.
Determine the population and sampling plan: Define who or what is being studied and how cases will be selected.
- Target population: The complete group about which conclusions are intended.
- Sampling frame: The operational list from which the sample is drawn, such as a hospital employee register.
- The plan states the sampling method, eligibility criteria, and desired sample size.
Develop instruments and conduct a pilot study: Prepare questionnaires, interview schedules, observation forms, tests, or measurement devices.
- Pilot study: A small preliminary administration detects ambiguous questions, impractical procedures, and timing problems.
- Reliability and validity evidence should be established for important scales.
Collect data: Apply the approved procedures consistently and ethically.
- Researchers obtain consent, protect confidentiality, record non-response, and maintain secure data files.
- Standard instructions reduce differences caused by how data collectors administer an instrument.
Process and analyse data: Inspect, organise, code, and evaluate the evidence.
- Quantitative analysis: May use frequencies, means, correlations, confidence intervals, regression, or hypothesis tests.
- Qualitative analysis: May involve transcription, coding, category development, thematic analysis, and interpretation of context.
- Missing values, unusual observations, and data-entry errors should be addressed transparently.
Interpret findings and draw conclusions: Relate results to the objectives, hypotheses, theory, and prior research.
- Statistical significance does not automatically imply practical importance; an effect must also be assessed by its size and context.
- Conclusions must remain within the design's limits. Correlation in a cross-sectional survey, for example, does not by itself prove causation.
Report and communicate the research: Present the problem, literature, methods, results, interpretation, limitations, and conclusions coherently.
- Tables and figures should identify variables, units, and sample sizes.
- Ethical reporting includes acknowledging contradictory evidence and avoiding fabrication, falsification, or selective presentation.

III. Research Design — The Blueprint of Inquiry

A. Research Design - Introduction and Types of Research Designs

Research design is the overall plan connecting a research question to the evidence needed for a credible answer.

Core function: It determines what data will be collected, from whom, when, under what conditions, and through which analytical procedures.
Exploratory design: Investigates a poorly understood issue to generate insights, concepts, or hypotheses.
- Flexible methods include literature exploration, expert interviews, focus groups, and case studies.
- Findings are usually provisional and are not intended to estimate population values precisely.
Descriptive design: Portrays the characteristics, frequency, or distribution of a phenomenon.
- A survey estimating the percentage of nurses experiencing high stress is descriptive.
- It answers questions such as what, who, where, and how much, but normally does not establish why a pattern occurs.
Correlational design: Measures the direction and strength of association between variables without manipulating them.
- A correlation coefficient ranges from (-1) to (+1); its sign shows direction and its magnitude shows linear strength.
- Confounding variables and reverse causation prevent association alone from proving causality.
Explanatory or causal design: Tests whether changes in one variable produce changes in another.
- A credible causal claim requires temporal order, covariation, and control of plausible alternative explanations.
Experimental design: Manipulates an independent variable, uses a comparison or control condition, and ordinarily assigns participants randomly.
- Random assignment helps distribute known and unknown confounders across groups.
- A pre-test/post-test experiment compares outcome changes before and after an intervention.
Quasi-experimental design: Examines an intervention without full random assignment.
- Examples include nonequivalent comparison-group and interrupted time-series designs.
- It is practical in schools, hospitals, and communities, but selection differences can weaken causal inference.
Cross-sectional design: Collects data at one point or over a short period.
- It efficiently estimates current characteristics but provides limited evidence about change or temporal sequence.
Longitudinal design: Collects repeated observations over time.
- Panel, cohort, and trend studies can examine development and ordering, although attrition may bias results.
Case-study design: Conducts an intensive investigation of a bounded case such as an organisation, programme, event, or community.
- Multiple evidence sources support contextual understanding, but statistical generalisation is usually limited.
Qualitative design: Uses forms such as phenomenology, ethnography, grounded theory, narrative inquiry, or qualitative case study.
- It emphasises meaning, experience, social process, and context through interviews, observations, documents, or artefacts.
Mixed-methods design: Integrates quantitative and qualitative evidence within one investigation.
- A sequential explanatory design may analyse survey results first and then use interviews to explain those results.

B. Design Quality and Limitations

A strong design maximises credible inference while acknowledging constraints that cannot be eliminated.

Internal validity: Indicates whether an observed effect can reasonably be attributed to the proposed cause rather than history, maturation, selection, or measurement changes.
External validity: Concerns whether findings apply to other populations, settings, and periods.
Construct validity: Evaluates whether operational measures adequately represent concepts such as stress, motivation, or achievement.
Trade-off: A tightly controlled laboratory experiment may have strong internal validity but weaker real-world applicability.
Practical constraint: Ethical rules may prohibit manipulation, while cost, attrition, inaccessible populations, or limited time may restrict design choices.

IV. Sampling — Selecting Cases for Investigation

A. Sampling

Sampling is the process of selecting a subset of units from a defined population so that evidence can be collected efficiently and, where justified, generalised.

Population and sample: A population contains all relevant units; a sample contains the units actually selected.
Parameter and statistic: A parameter, such as population mean (\mu), describes the population; a statistic, such as sample mean (\bar{x}), estimates it.
Probability sampling: Every population unit has a known, non-zero selection probability.
- Simple random sampling: Units are selected by a random mechanism from the sampling frame.
- Systematic sampling: After a random start, every (k)-th unit is chosen, where (k=N/n), (N) is population size, and (n) is sample size.
- Stratified sampling: The population is divided into strata, such as departments, and samples are drawn from each.
- Cluster sampling: Naturally occurring groups, such as hospitals or classes, are sampled before individuals are studied.
Non-probability sampling: Selection probabilities are unknown.
- Convenience sampling: Includes readily accessible cases.
- Purposive sampling: Selects information-rich cases using stated criteria.
- Quota sampling: Fills predefined category totals without random selection.
- Snowball sampling: Existing participants recruit others, which is useful for hard-to-reach populations.
Sample size: Required size depends on population variability, desired precision, confidence level, design, expected effect, and non-response.
Sampling error: Random samples differ from the population by chance; larger well-designed samples generally reduce this error.
Sampling bias: Systematic distortion occurs through incomplete frames, undercoverage, self-selection, or non-response and is not automatically corrected by increasing sample size.

B. Applications and Limitations

Appropriate sampling balances representativeness, precision, access, ethics, and available resources.

Quantitative application: Probability samples support estimation of population values and calculation of sampling uncertainty.
Qualitative application: Purposive sampling prioritises relevance, variation, and conceptual depth rather than statistical representativeness.
Representativeness: It depends on selection quality and response patterns, not merely on the number of participants.
Generalisation limit: Findings from one convenient group cannot defensibly be extended to an entire population without strong supporting evidence.
Documentation: Researchers should report the population, frame, method, eligibility criteria, sample size, response rate, and reasons for exclusions.