Research is a systematic, controlled, empirical, and critical investigation of natural or social phenomena, guided by theory and hypotheses, that aims to discover new facts, verify existing knowledge, or establish new relationships. The word comes from the French recherche — to search again, implying a thorough and iterative inquiry.
Clifford Woody summarised it succinctly: "Research comprises defining and redefining problems, formulating hypotheses or suggested solutions; collecting, organising and evaluating data; making deductions and reaching conclusions; and at last carefully testing the conclusions to determine whether they fit the formulating hypothesis."
In statistics, research specifically implies the quantitative study of variability — designing experiments or surveys that yield numerical data, analysing them with appropriate statistical methods, and drawing inferences that are supported by probabilistic reasoning.
Every research project is driven by one or more of the following objectives:
The National Sample Survey Office (NSSO) conducts household consumption surveys to describe the average monthly per-capita expenditure across Indian states. The objective is descriptive — to estimate population parameters (mean expenditure, its variance, its distribution) — not to explain why some states are richer.
A researcher hypothesises that gender discrimination affects salary levels in the private sector. The objective is explanatory — to test whether, after controlling for education and experience, gender has a statistically significant effect on salary. A regression model with salary as the dependent variable and gender (dummy), education, and experience as predictors is fitted. A significant negative coefficient for the gender dummy supports the hypothesis.
Research is motivated by a mix of intellectual curiosity and practical need. Kothari (2009) identifies several categories:
A mathematician studies the distribution of prime numbers — there is no immediate application, but the work advances human knowledge and may eventually enable cryptographic breakthroughs. This is pure (basic) research, driven solely by curiosity.
A pharmaceutical company needs to determine the optimal dosage of a new drug. A randomised controlled trial is designed with three dose levels and a placebo. The research is motivated by a practical need — obtaining regulatory approval — and the results will directly influence the drug's labelling. This is applied research.
| Type | Goal | Example |
|---|---|---|
| Descriptive | Portray accurately the characteristics of a situation or group | Census of India — describing population demographics |
| Analytical (Explanatory) | Understand cause-and-effect relationships using existing data | Analysing why dropout rates are higher in rural schools |
| Predictive | Forecast future events based on current data and models | Time-series forecasting of stock prices |
| Type | Approach | Key Feature |
|---|---|---|
| Quantitative | Measurement and statistical analysis of numerical data | Hypothesis testing, confidence intervals, p-values |
| Qualitative | In-depth understanding via interviews, observation, narratives | Thematic analysis, case studies, grounded theory |
| Mixed methods | Combines both quantitative and qualitative approaches | Triangulation of findings for richer conclusions |
| Type | Source | Example |
|---|---|---|
| Primary | Data collected first-hand by the researcher | Field survey, laboratory experiment |
| Secondary | Pre-existing data collected by others | Census reports, National Family Health Survey data |
| Type | Design | Example |
|---|---|---|
| Cross-sectional | Observation at one point in time | A survey of student anxiety levels in December 2024 |
| Longitudinal | Observations over an extended period | Tracking the same cohort of students' anxiety over 4 years |
| Type | Setting | Example |
|---|---|---|
| Laboratory / Experimental | Controlled environment | Testing crop yield under controlled fertilizer levels |
| Field / Non-experimental | Natural setting | Observing consumer behaviour in a real market |
A researcher sends questionnaires to 500 households in Kadapa district to study the relationship between household income and expenditure on education. This is: descriptive (describing the relationship), quantitative (income and expenditure are numerical), primary (researcher collects the data), cross-sectional (one-time survey), and field-based (conducted in respondents' homes).
A health researcher follows 200 diabetic patients for 5 years, measuring blood glucose levels (quantitative) every 6 months and conducting in-depth interviews about diet and lifestyle (qualitative) annually. This is a mixed-methods, longitudinal, primary research design that combines the statistical power of quantitative analysis with the depth of qualitative insight.
Starts with a theory → derive a hypothesis → collect data to test the hypothesis → confirm or refute the theory. This is the dominant approach in quantitative research.
Example: Theory: "Inflation reduces purchasing power." Hypothesis: "A 5% increase in CPI reduces real consumer spending by 2%." Test with time-series data.
Starts with observations → identify patterns → formulate a tentative hypothesis → develop a theory. This is common in qualitative and exploratory research.
Example: Observing that students who attend tutorials regularly score higher → hypothesising that tutorial attendance improves exam performance → developing a theory about structured peer-learning.
Theory: "Employee satisfaction reduces turnover." Hypothesis: "Companies with satisfaction scores above 75 have annual turnover below 10%." Collect satisfaction survey data and HR turnover records from 30 companies. Compute Pearson's correlation and fit a regression. If the slope is significant and negative, the hypothesis is supported. This is deductive — we started with a theory and tested it.
A researcher interviews 20 victims of workplace harassment without a pre-set hypothesis. Through thematic analysis, three recurring themes emerge: fear of retaliation, lack of institutional support, and power imbalance. These themes generate a tentative theory: "Workplace harassment persists because institutional responses are perceived as ineffective." This theory can now be tested deductively in a larger survey.
A research problem is a clear, specific, and answerable question or set of questions that the researcher intends to resolve through systematic investigation. It is the starting point of every research project — without a well-defined problem, the entire study lacks direction.
Problem: "Is there a significant difference in the average monthly salary of male and female employees in the IT sector in Bengaluru, after controlling for years of experience and education level?"
This problem is: clear (specific population, variables, and comparison), feasible (salary surveys exist), relevant (gender pay gap is a policy issue), and testable (use ANCOVA or regression with a gender dummy).
Problem: "What is the effect of unemployment on society?"
This is too broad — "effect" could mean economic, psychological, health, crime, political, and many other outcomes. "Society" is unbounded. The problem needs to be narrowed: "What is the effect of long-term unemployment (>12 months) on self-reported mental health (PHQ-9 score) among adults aged 25–45 in one city?" Now it is specific, measurable, and testable.