Research Aptitude — Meaning, Types & Methods

Research is classified on several independent axes at once, and this is the single most useful thing to understand before memorising anything. A study can be applied and quantitative and experimental all at the same time, because those three labels answer three different questions: what is the study for (purpose), what form does its data take (nature), and how is it carried out (method). Questions that look tricky are usually just mixing axes.

Two more axes appear occasionally. On time: a cross-sectional study takes one snapshot of different age groups at a single moment, while a longitudinal study follows the same subjects over years — deeper, but expensive and prone to sample mortality. On number of investigators and disciplines: inter-disciplinary research fuses the concepts of two or more disciplines into a single framework, whereas multi-disciplinary research places disciplines side by side, each keeping its own tools.

A paradigm (Thomas Kuhn’s term) is the set of shared assumptions a research community holds about what reality is (ontology), how it can be known (epistemology) and therefore how it should be studied (methodology). NET asks about paradigms because they explain why quantitative and qualitative researchers argue: they are not disagreeing about a technique, they are disagreeing about the nature of truth.

• Positivism (Auguste Comte). There is one objective reality, external to the observer and governed by law. It can be measured directly and value-free. Method: quantify, hypothesise, verify, generalise; the researcher stays a detached outsider. Preferred designs: experiment and large survey. The model of enquiry is borrowed from the natural sciences and applied unchanged to society.
• Post-positivism (associated with Karl Popper’s falsificationism and critical realism). Reality does exist, but our knowledge of it is always imperfect, probabilistic and theory-laden. Observation is never perfectly neutral, so a hypothesis can never be finally proved — only falsified or provisionally supported. Method stays largely quantitative but becomes modest: multiple measures, triangulation, confidence levels instead of certainty.
• Interpretivism / constructivism (roots in Max Weber’s verstehen and in phenomenology). There is no single reality; social reality is constructed in the minds of participants, so there are as many realities as there are perspectives. The researcher must enter the actor’s frame of reference and interpret meaning, accepting that his own values are part of the process. Method: qualitative — depth interview, ethnography, case study, small purposive samples, inductive theory building.
• Critical / transformative paradigm. Knowledge is never neutral; research should expose power and produce emancipation (feminist, Marxist, Dalit and participatory traditions). Pragmatism, the fourth commonly listed paradigm, refuses the quarrel altogether: use whatever mix of methods answers the question — it is the philosophical home of mixed-methods research.

The exam asks two kinds of questions here: "which step comes first?" and "arrange in correct sequence". Both are settled by one insight — the literature review comes after the problem is identified but before the hypothesis is framed. You cannot review literature on nothing, and you cannot responsibly guess an answer before you know what others have found. Learn the chain in that order and the sequencing questions become free marks.

1. Identification and formulation of the problem — move from a broad area ("teacher stress") to a researchable statement of the problem with clear scope, and define terms operationally (state exactly how each concept will be measured).
2. Review of related literature — find what is already known, avoid duplication, borrow tested tools, and above all locate the research gap that justifies your study. It also prevents you from re-inventing a wheel that turned out to be square.
3. Formulation of objectives and hypotheses — convert the gap into testable statements. Purely exploratory and most qualitative studies may legitimately have research questions instead of hypotheses.
4. Preparing the research design — the blueprint: which type of study, which variables, which comparison groups, what time frame, what budget. A design is judged by how well it removes alternative explanations.
5. Defining the population and drawing the sample — decide the universe, the sampling frame, the technique and the size.
6. Constructing and standardising the tools, then collecting the data — pilot study and pre-testing belong here; they catch a bad item before it ruins 500 responses.
7. Analysis of data — editing, coding, classification and tabulation first; then descriptive statistics and the appropriate test of significance.
8. Interpretation, generalisation and hypothesis testing — statistics give numbers, interpretation gives meaning. Say plainly what the result implies and what it cannot claim.
9. Reporting — the thesis, paper or report, with limitations honestly stated and suggestions for further research. Research that is not communicated has not finished.

A hypothesis is a tentative, testable statement about the expected relationship between two or more variables — an intelligent guess put in a form that evidence can knock down. It is not a wish, not a question, and not a conclusion. Its whole value lies in being falsifiable: if no possible observation could contradict it, it is not a hypothesis.

• Null hypothesis (H0) — states that there is no significant difference / no relationship ("there is no significant difference in achievement between students taught by lecture and by discussion"). This is the hypothesis that is actually put to statistical test, because "no difference" is a precise claim while "some difference" is not.
• Alternative hypothesis (H1 / Ha) — the researcher’s own expectation, accepted only if H0 is rejected. It may be non-directional ("there is a significant difference" — two-tailed test) or directional ("group A will score higher than group B" — one-tailed test). A directional hypothesis needs a strong theoretical reason, because it commits you to one side in advance.
• Research (substantive) hypothesis — the statement in ordinary language, before it is dressed in statistical form; statistical hypothesis is the same claim expressed in terms of population parameters.
• Characteristics of a good hypothesis: conceptually clear and free of ambiguity; empirically testable; specific and limited in scope; consistent with the known body of theory; stated in the simplest terms that will do; and framed before the data is examined.

A variable is any characteristic that can take more than one value. The four names the exam insists on are these. The independent variable is the presumed cause, the one manipulated or classified by the researcher (teaching method). The dependent variable is the presumed effect, the one measured as outcome (achievement score). An extraneous (or nuisance) variable is an outside variable that could also affect the dependent variable and so must be controlled — if it is not controlled and it actually distorts the result, it has become a confounding variable. An intervening (mediating) variable is the invisible internal process that links independent to dependent and is inferred rather than measured directly — motivation, anxiety, fatigue.

Two more appear in better papers. A moderator variable changes the strength or direction of the independent-dependent relationship (a method works well for boys, poorly for girls: gender moderates). A control variable is an extraneous variable that the researcher has deliberately held constant — testing only students of one age group, for example. Keep the logic straight: extraneous describes a threat, control describes what you did about it, confounding describes the threat that got away.

The population (or universe) is every unit you want to speak about; the sample is the part you actually study; the sampling frame is the list from which the sample is drawn — the electoral roll, the school register. The gap between the sample result and the true population value is sampling error, and it shrinks as the sample grows. Errors that come from a bad frame, careless interviewing or non-response are non-sampling errors, and these do not shrink with size — a bigger flawed study is simply a bigger flaw. Census studies every unit; sampling studies a part and infers.

The great divide: in probability sampling every unit of the population has a known, non-zero chance of selection, which is exactly what licenses you to generalise and to calculate a margin of error. In non-probability sampling selection depends on the researcher’s judgement or convenience, the chance of selection is unknown, and therefore statistical generalisation is not justified — which does not make it bad, only differently useful.

• Questionnaire — a set of written questions filled in by the respondent himself and usually sent by post or online. Cheap, wide reach, respondent stays anonymous; but low return rate, useless for illiterate respondents, and no chance to clarify a misunderstood item.
• Schedule — the same kind of form, but filled in by the investigator or enumerator while asking the respondent face to face. Higher response rate, usable with illiterate respondents, doubts can be cleared; but costly, slow, and open to interviewer bias. This one-word difference — who holds the pen — is the whole answer to a very common NET question.
• Interview — oral, face-to-face or telephonic. Structured (fixed questions, fixed order, comparable answers), unstructured (a guide only, free flow, rich but hard to compare), semi-structured, focus group (six to twelve participants discussing together so that ideas spark off each other), and depth interview for motives and feelings.
• Observation — the only tool that records behaviour as it happens, without depending on what people say about themselves. Participant (the observer joins the group) versus non-participant; structured (a prepared schedule and coding) versus unstructured; controlled (laboratory) versus naturalistic (field). Its weaknesses are observer bias and the fact that attitudes and past events cannot be observed at all.
• Scales and standardised tests — for attitudes and traits that cannot be counted directly. Likert scale (a statement rated on a five- or seven-point agree-disagree continuum, scores summed); Thurstone scale (judges assign weights to statements in advance); Guttman scale (cumulative — agreeing with an item implies agreeing with all weaker items); semantic differential (rating between two opposite adjectives); sociometry (mapping choices within a group). Also used: rating scales, checklists, inventories, projective techniques and document analysis.

Reliability is consistency; validity is truthfulness. A reliable tool gives the same result on repetition under similar conditions. A valid tool measures what it claims to measure and nothing else. The relationship between them is the point the exam tests: a tool can be perfectly reliable and completely invalid, but a valid tool is necessarily reliable. In short, reliability is a necessary but not sufficient condition for validity.

• Methods of establishing reliability: test-retest (same test, same people, later occasion — gives coefficient of stability); parallel / equivalent forms (two matched versions — coefficient of equivalence); split-half (score odd items against even items and correlate, corrected by the Spearman-Brown formula); internal consistency (Cronbach’s alpha, Kuder-Richardson); and inter-rater reliability for observation and essay marking.
• Kinds of validity: face validity (does it merely look right to a layman — the weakest); content validity (does it cover the whole syllabus or domain, judged by experts — vital for achievement tests); construct validity (does it really tap the abstract trait, shown by convergence with related measures and divergence from unrelated ones); and criterion-related validity, which splits into concurrent (agrees with an accepted measure taken at the same time) and predictive (forecasts a future performance, as an entrance test predicts college grades).
• Internal versus external validity of a design (not of a tool): internal validity asks whether the observed change was really caused by the independent variable and not by history, maturation, testing effect or selection bias; external validity asks how far the finding can be generalised to other people, places and times. Tightening controls usually raises internal validity and lowers external validity — the laboratory-versus-field trade-off.
• Objectivity and usability complete the set: a tool is objective if different scorers award the same marks to the same answer, and usable if it is practical in cost, time and ease of administration. In qualitative research the parallel criteria are credibility, transferability, dependability and confirmability, often supported by triangulation — checking a finding against more than one method, source, investigator or theory.