Unit 2: Research Aptitude mind map

Meaning of research, the scientific method, theory, and the philosophies behind research.

What research is

Meaning of research

Research is a careful and systematic search for knowledge, using evidence and reasoning.

• Systematic, logical and based on evidence.
• Challenges the status quo.
• Uses both deductive and inductive reasoning.
• A critical and passionate search for knowledge.

Research is more than collecting facts. It asks a question, gathers evidence in a planned way and draws a conclusion that others can check.

Research uses two kinds of reasoning together. It goes from theory to data (deduction) and from data to theory (induction).

Merely proving one's own belief, or only reviewing what others wrote, is not the meaning of research.

Characteristics of scientific research

Scientific research is objective, systematic, accurate and checkable by others.

• Empirical: based on observation and evidence.
• Systematic, accurate and precise.
• Objective, not subjective.
• Verifiable by other researchers.
• Done under controlled conditions.

Scientific observations rest on verifiable evidence. Anyone should be able to repeat the study and check the result.

Scientific research excludes subjectivity, unethical practices and non-predictive inferences. Uncontrolled conditions are not a mark of scientific research.

What research is not

Some statements about research look true but are false.

• Common sense and personal beliefs are not research.
• Research does not always build a new theory.
• Hypothesis testing is not done in every study.
• Research does not always improve the area it studies.

Common sense is built from accumulated everyday experience. It is not checked in a systematic way, so it is not research.

Qualitative studies often do not test a hypothesis. Findings may also never be used, so research does not necessarily improve practice.

Kerlinger's four ways of knowing

Kerlinger named four ways in which people reach answers to questions.

• Method of tenacity: holding on to a belief.
• Method of authority: accepting what experts say.
• Method of intuition (a priori): what seems self-evident.
• Method of science: testing against evidence.

The first three can be wrong because the belief is never tested. Only the method of science checks a claim against observation.

Creativity and non-functionality are not among the four ways.

Four operations behind a cause-and-effect claim

To claim that one thing causes another, research must do four things.

• Show that the two things vary together (co-variance).
• Remove spurious (false) relations.
• Show the correct order in time.
• Offer a theory that explains the link.

Co-variance means that when one changes the other changes. A spurious relation is a link caused by a third factor, so it must be ruled out.

The cause must come before the effect, so time order matters. A theory then explains why the cause leads to the effect.

Theory and the scientific method

Concept, construct and variable

A concept is an idea. A variable is the form of that idea that can be measured.

• Concept: an idea such as intelligence or achievement.
• Construct: a concept created on purpose for research.
• Variable: the empirical counterpart of a concept.
• A variable takes different values.

You cannot measure 'intelligence' directly. You turn it into a variable, such as a test score, and then measure that.

Theory, model and survey

A theory explains and predicts. A model shows a system in a simple form.

• Theory: connected statements that explain and predict a phenomenon.
• Model: a simplified representation of a system or object.
• Survey: a way of collecting data, not a theory.
• A theory generates hypotheses that can be tested.

Theory and data support each other. A theory suggests hypotheses. The data then confirm or reject them, and the theory is revised.

A sound theory and middle range theory

A good theory is simple, testable and fits what is already known.

• Stated simply and precisely.
• Can be verified.
• Guides new discoveries.
• Conforms to established knowledge.
• Middle range theories work in a limited domain.

A theory that does not fit validated earlier knowledge is not a sound theory. It must not ignore what has been established.

Merton proposed middle range theories. They sit between grand theories that try to explain everything and plain observation. They explain a limited set of events and can be tested.

Steps of the scientific method

Scientific inquiry follows a fixed order from the problem to the verified result.

• Identify the problem.
• Formulate a hypothesis.
• Collect, organise and analyse data.
• Interpret the results.
• Verify, reject or modify the hypothesis.

The scientific method is open to criticism, uses clear concepts, and aims at general laws or theories. It is committed to objective considerations only.

A statement that says the method is 'not committed to objective considerations' is a wrong postulate.

Deduction and induction

Deduction goes from theory to data. Induction goes from data to theory.

• Deduction order: theory, hypothesis, data collection, findings, hypothesis confirmed or rejected, revise theory.
• Induction order: specific observations, pattern, hypothesis, data, analysis, general theory.
• Deduction tests a theory.
• Induction builds a theory.

A deductive study starts with a general idea and tests it. An inductive study starts with cases and arrives at a general idea.

'This swan is white, that swan is white, so all swans are white' is induction. It also shows the danger, since the conclusion goes beyond the evidence.

Drawing general conclusions from observations is induction. Do not call it deduction.

Inductivism, deductivism and the hypothetico-deductive model

Inductivism builds laws from facts. Deductivism tests laws through hypotheses.

• Inductivism: gather facts first, laws come from the facts.
• Deductivism: theory gives a hypothesis, the hypothesis is tested.
• The hypothetico-deductive order: observation, hypothesis, data collection, analysis.
• Empirico-inductive paradigm: ethnography, case study, narrative methods.

When a question says 'knowledge is arrived at through gathering facts that give the basis for laws', the answer is inductivism.

When it says 'theory is used to generate hypotheses that are then tested', the answer is deductivism.

Philosophy behind research

Ontology, epistemology, methodology and method

These four words describe different levels of a research approach.

• Ontology: what social reality is.
• Epistemology: what counts as knowledge and how we know it.
• Methodology: the overall approach to study.
• Method: the actual tool, such as a survey.

Realism is an ontological position. Positivism is an epistemological position. The ideographic approach is a methodology. A survey is a method.

The question 'can the social world be studied with the methods of natural science' is an epistemological question.

Positivism and empiricism

Positivism says that social life can be studied with the methods of natural science.

• Knowledge comes from observation and measurement.
• Associated with quantitative research and data-based research.
• August Comte is the founder.
• Empiricism: only knowledge gained through the senses is accepted.

Positivism stresses objectivity, measurement and testing. Its supporters use the same logic that the natural sciences use.

Interpretivism is the opposite view. It is therefore not a principle of positivism.

Interpretivism and constructionism

Interpretivism says that people must be understood through the meanings they give to their actions.

• Interpretivism: the epistemology of qualitative research.
• Constructionism: the ontology of qualitative research.
• Social reality is built and rebuilt by people.
• Max Weber is linked with interpretation (verstehen).
• Objectivism says social reality exists apart from people.

Constructionism says that social phenomena and their meanings are continually being accomplished by social actors. Reality is not fixed.

Objectivism says the opposite. It is the ontology of quantitative research.

Relativism and realism

Relativism says behaviour must be understood within its own culture.

• Relativism: no universal standard of judgement.
• Absolutism and universalism say the opposite.
• Realism: reality exists independent of our views.
• Realism has empirical and critical forms.

When a question says that behaviour can be understood only in the context of the culture where it occurs, the answer is relativism.

Qualitative researchers criticise quantitative research for reductionism. That means breaking a rich reality into a few numbers.

Popper, Kuhn, Feyerabend and Bacon

Four thinkers shaped how we think about science.

• Karl Popper: falsification, a theory must be open to being proved wrong.
• T.S. Kuhn: paradigm and paradigm shift.
• Paul Feyerabend: 'Against Method', no single scientific method.
• Francis Bacon: the inductive method.

Popper said a claim is scientific only if some observation could prove it wrong. A claim that nothing can disprove is not scientific.

Kuhn said science works inside a paradigm for a long time. When too many problems pile up, the paradigm shifts.

Post-positivism and postmodernism

Post-positivism softens positivism. Postmodernism rejects fixed truth.

• Post-positivism: knowledge is never perfectly certain.
• Postmodernism rejects fixed notions of reality.
• Postmodernism resists certainty and closure.
• Postmodernism does not accept grand narratives as the base of research.

Post-positivism still values evidence, but it accepts that the researcher cannot be fully neutral. It gives a place to qualitative work.

Postmodernism looks for subjective understanding. A statement saying that it accepts all grand narratives is the incorrect one.

The kinds of research, the steps in order, the problem, the literature review and the hypothesis.

Types of research

Basic and applied research

Basic research adds to knowledge. Applied research solves a practical problem.

• Basic (pure, fundamental) research: knowledge for its own sake, builds theory.
• Applied research: solves a practical problem.
• Example of basic: development of the theory of relativity.
• Example of applied: comparing two teaching methods on achievement.

Basic research has no immediate use in mind. It tests or expands theory. Applied research takes a problem from real life and looks for a solution.

Comparing the effect of two teaching methods on student achievement is applied research, because the result can be used in classrooms.

Action research

Action research helps a practitioner improve their own work in their own setting.

• Cyclic: plan, act, observe, reflect.
• Solves a local problem.
• Does not aim to build theory.
• Does not aim to generalise findings.
• No random sampling and no strict control of variables.

A teacher who wants to improve the study habits of their own class uses action research. The teacher plans a change, tries it, watches the result and reflects. Then the next cycle starts.

Because of this, action research develops the professional expertise of the researcher-practitioner.

Evaluation and formative research

Evaluation research judges a programme. Formative research improves it while it is being made.

• Evaluation research: assesses programmes or interventions.
• Formative research: done while a programme or product is being developed.
• Focus group discussion is a formative research method.
• Formative research can be qualitative and quantitative.

Evaluation research asks how well a programme works. Formative research asks how to make it better before it is finished.

Exploratory, descriptive and explanatory research

These three types differ in how much is already known and what the study wants.

• Exploratory: little or no information, builds first understanding.
• Descriptive: describes social situations, events, systems and structures.
• Explanatory: explains why something happens, the causes.
• 'Why does stressful living lead to heart attack?' is explanatory.

When a topic is new, the researcher first explores it. Then the researcher describes what exists. Later the researcher explains the causes.

A study on how many crimes women commit is descriptive. A study on why they commit crimes is explanatory.

Experimental, ex post facto and correlational research

These types differ in whether the researcher controls the cause.

• Experimental: control, manipulate and observe.
• Ex post facto: studies a cause that has already happened.
• Correlational: finds whether variables vary together.
• Experimental research gives the strongest evidence of cause and effect.

In an experiment the researcher changes the independent variable. In ex post facto research the researcher cannot change it, so existing data are used to make groups for comparison.

Studying how parents' socio-economic status affects the drop-out rate is a non-experimental hypothesis testing study. The researcher cannot manipulate status.

Normative, feminist and interdisciplinary research

Some research types are named by their purpose or their perspective.

• Normative research: raises the issue so that the conclusion is implied.
• Feminist research: focuses on inequality between men and women.
• Interdisciplinary research: uses more than one subject.
• Interdisciplinary aim: a holistic approach.

Normative research deals with what 'ought' to be. The answer is already built into the way the question is asked.

Interdisciplinary research looks at a problem through ideas and methods from several subjects, to see the whole problem and not just one side.

Steps of the research process

The order of the research process

Research moves step by step from a problem to a report.

• Choose a topic and define the problem.
• Review the literature.
• Formulate the hypothesis or question.
• Design the research.
• Collect data, analyse and interpret.
• Report the findings.

Different books show slightly different lists, so read the options carefully. The usual order is problem, literature review, design, data collection, analysis and report.

In quantitative research, a common order is hypothesis, research design, process data, analyse data and findings.

Choosing and finalising the research problem

A good problem is chosen after reading, and is original and workable.

• Select the area of research.
• Study earlier reports and reference material.
• Find the research gap.
• Identify and define the problem.
• Originality needs uniqueness, usefulness, support and feasibility.

The formulation of a research problem is like laying the foundation of a building. Everything else rests on it.

Not needing a supervisor has nothing to do with originality.

Formulating the research problem

Formulating a problem means stating the aim, the specific questions and the conceptual model.

• State the object of the study.
• State the specific questions.
• State the conceptual model.
• State how hypotheses will be built.
• Research questions define the field: what will be studied and how far.

A research question gives the study its direction. It fixes the field of the study and its limits.

Review of literature

A literature review studies what earlier researchers have found.

• Gives subject background.
• Helps to form research questions.
• Shows methods used earlier.
• Avoids repeating earlier work.
• Helps in interpreting findings.

A review is an integral part of the research process. The statement that it is not integral is false.

It does not debate the ethics of research, and it does not give direct answers to your question.

A systematic review follows explicit procedures. It states how studies were searched and selected, so that another person can repeat it.

Research gaps and steps of a literature review

A research gap is a place where knowledge is missing or unclear.

• Under-researched areas are gaps.
• Areas with inconsistent findings are gaps.
• Review steps: key terms, locate, evaluate, organise, write.
• Spotting gaps is the chief way of finding research questions.

A gap can be a topic nobody has studied. It can also be a topic where earlier studies disagree. Both give you a question to work on.

Differences in the backgrounds of researchers are not a gap in knowledge.

Hypothesis: meaning and qualities

A hypothesis is a tentative, testable statement about a relationship.

• Tentative explanation, yet to be tested.
• Simple, clear and specific.
• States the relation between variables.
• Describes one issue only and is empirically testable.
• Not a 'proved assumption'.

A hypothesis is a smart guess based on theory or earlier findings. After testing, it is accepted or rejected.

A statement with 'should' is advice, not a hypothesis. 'Eating 20 almonds a day lowers the risk of heart disease' can be tested, so it is a valid hypothesis.

Sources of hypotheses and the creative step

Hypotheses come from earlier research, beliefs and the researcher's own experience.

• Reports of earlier research.
• Beliefs in the area of the problem.
• Prior experience of the researcher.
• Sampling method and type of population are not sources.
• Most creativity: identifying variables and hypothesising.

After the variables are identified and the theory is set, the next step is to generate the hypothesis. This step needs imagination, because the researcher must think how things are related.

Quantitative and qualitative approaches

What quantitative research is concerned with

Quantitative research is structured and uses numbers. Its main concerns are four.

• Measurement.
• Causality.
• Generalisation.
• Replication.
• Deductive approach to theory.

Quantitative research views social reality as an external, objective reality that can be measured. It begins with a theory and tests it with data.

Replication means repeating the study to check the result. Emphasis on context is not a quantitative concern.

What qualitative research is concerned with

Qualitative research seeks meaning in context. It is flexible.

• Description.
• Emphasis on context.
• Emphasis on process.
• Flexibility and limited structure.
• Inductive approach to theory.

Qualitative research sees social reality as constantly changing, shaped by people. Its questions are not a fixed set, so they can change as the study goes on.

Generalisation is a quantitative concern, so it is a wrong option here.

Quantitative and qualitative compared

The two differ in structure, theory and view of reality.

• Quantitative: highly structured, standardised questions, deductive.
• Qualitative: unstructured, flexible questions, inductive.
• Quantitative: objectivism and positivism.
• Qualitative: constructionism and interpretivism.

Quantitative research is also called the structured approach. Qualitative research is called the unstructured approach.

Qualitative researchers criticise quantitative research for reductionism, because it reduces rich social life to a few numbers.

What the two approaches share

Both approaches have common aims despite their differences.

• Both answer research questions.
• Both are concerned with data reduction.
• Both are concerned with variation in data.
• Both give importance to transparency.

Data reduction means making large data manageable. Transparency means showing clearly how the work was done.

Emphasis on contextual understanding belongs mainly to qualitative research. So it is not a shared element.

Experimental, observational, qualitative and historical methods, and what each is best at.

Experimental and observational methods

Experimental method

In an experiment the researcher changes one thing, controls the rest and watches the effect.

• Gives the strongest evidence of cause and effect.
• Manipulates the independent variable.
• Removes the influence of extraneous factors.
• Stages: setting, design, variables, administer, analyse.

The order of stages is: select the setting, select the design, operationalise the variables, administer the experiment, analyse and interpret.

Correlation and description cannot give such strong evidence of cause and effect.

Control in experiments

Control means stopping other factors from spoiling the result.

• Control the physical setting.
• Choose participants carefully.
• Use statistical analysis.
• Give the treatment the same way every time.
• Pilot test, revise and test again.

Random error is unpredictable and changes from one measurement to another. Giving the treatment in the same way each time reduces it.

Test-retest reliability shows how free a measure is from random error. If the same people score almost the same twice, error is small.

Experimental designs and the placebo group

Designs differ in the groups used and in when the test is given.

• Placebo: a fake treatment that looks real.
• A placebo design with a control group has three groups.
• Informal designs: before-and-after without control, after-only with control, before-and-after with control.
• Factorial design: two or more independent variables, one dependent variable.

The three groups are the experimental group, the placebo group and the control group. The control group gets nothing.

Factorial designs and randomised block designs are formal designs, so they are not informal designs.

Field experiment and its limits

A field experiment is done in a natural setting, so people behave naturally.

• Strong external validity.
• Subjects behave normally.
• Natural setting.
• Gives primary data.
• Laboratory criticism: artificial conditions.

Because a field experiment takes place in real life, its results apply better to real life. That is strong external validity.

A valid criticism of laboratory experiments in the social sciences is the artificiality of laboratory conditions.

Hawthorne effect, reactivity and blinding

People may change their behaviour just because they know they are studied.

• Hawthorne effect: behaviour improves because of attention.
• Subject reactivity: acting differently when monitored.
• Blind study: participants do not know their group.
• Blind review: the reviewer does not know the author.

When subjects improve not because of the experimental change but because they feel watched, it is the Hawthorne effect.

In a blind study, participants cannot guess what the experimenter wants. This protects the result.

Observational research and participant observation

Observational research watches behaviour in its natural setting, without changing anything.

• No manipulation of variables.
• Participant observation: the researcher joins the group.
• The researcher acts as a 'professional stranger'.
• Observation schedule lists the categories of behaviour.

Manipulation and control belong to experiments. In observation the researcher only watches and records.

A professional stranger is close enough to understand the group, and distant enough to stay objective.

Structured observation and observer errors

Structured observation uses a fixed list of behaviours, so different observers record the same way.

• Focal sampling: one person observed for a set time.
• Inter-observer reliability: agreement between observers.
• Halo effect: one good impression colours all ratings.
• Observation schedule: the list of behaviours to record.

High agreement between observers means the result does not depend on who is watching.

If an observer rates a person high on one aspect and lets that rating lift other ratings, that is the halo effect.

Ex post facto and field research

Ex post facto research studies a cause after it has already acted.

• Probes into causal factors from observed effects.
• Uses existing data to form groups.
• Field research collects data in real-life settings.
• Personal interview is a field research method.

In ex post facto research, the researcher does not manipulate anything. The researcher looks back from the effect to the likely cause.

Analysing documents and old records is not typical field research.

Qualitative methods

What qualitative research looks like

Qualitative research uses words, not numbers, and the researcher is the main tool.

• Researcher as key instrument.
• Data are words from a small number of people.
• Inductive data analysis.
• Context-specific generalisation.
• Avoids positivist assumptions.

Qualitative studies state the purpose and questions in a broad way. They use text analysis to interpret the larger meaning of the findings.

Representative quotations from participants support the findings. Scaling and SPSS belong to quantitative research.

Strengths and criticisms of qualitative research

Qualitative data are rich, but the method is hard to repeat.

• Data come from a sensitive social, historical and time context.
• Too many details of the setting.
• Hard to replicate.
• Context sensitivity cannot be completely removed.
• Survey method is not a qualitative method.

Critics say that qualitative studies give too many details and are hard to repeat. These are true statements.

Thematic analysis, case study and discourse analysis are qualitative. The survey method collects data from a large sample in numbers, so it is quantitative.

Case study

A case study is an in-depth study of one case, such as a person, a school or an institution.

• Particularistic, descriptive and inductive.
• Bounded system.
• Natural setting and meanings given by people.
• Hard to generalise to the whole population.

A case study uses many sources of data to understand the case fully. It is not simulated in a laboratory.

Its main difficulty is that findings are hard to generalise to the population at large.

Ethnography and auto-ethnography

Ethnography studies a group by living among its members.

• Researcher gets immersed in the social setting for a long time.
• Listens to conversation, observes behaviour.
• Collects documents and holds interviews.
• Auto-ethnography analyses the researcher's own personal experience.

The ethnographer does not use rigid structured interviews. The aim is to understand the way of life of the group from inside.

Studying peer interaction in a racially mixed classroom is an example of ethnography.

Grounded theory

Grounded theory builds a theory from data, step by step.

• Begins with a general research question.
• Theoretical sampling until theoretical saturation.
• Tool: constant comparison.
• Outcomes: concepts, categories and hypotheses.
• Does not use statistical procedures.

Order: general question, collect relevant data, code the data, saturate the categories, explore relations between categories.

Saturation means new data add no new ideas. A narrative is not an outcome of grounded theory. It is an outcome of narrative research.

Narrative research, phenomenology and hermeneutics

These three qualitative approaches look at stories, lived experience and interpretation.

• Narrative research: individual stories of people's lives.
• Phenomenology: meaning of lived experience.
• Hermeneutics: theory and method of interpretation, drawn from theology.
• Hermeneutics began with sacred texts.

Narrative research gathers stories through interviews and documents and puts them in sequence.

Analysing a person's life for meaning has the focus of phenomenology.

Steps in qualitative analysis

Qualitative analysis moves from raw data to a story.

• Organise and prepare the data.
• Code the data.
• Identify themes and patterns.
• Develop a storyline.
• Represent and interpret.

Inductive logic in a qualitative study goes from information about subjects, to questions, to themes and categories, to broad patterns, and then to a theory.

Trustworthiness in qualitative research

Qualitative research has its own criteria. Each one parallels a quantitative criterion.

• Credibility parallels internal validity.
• Transferability parallels external validity.
• Dependability parallels reliability.
• Confirmability parallels objectivity.

Transferability asks whether findings apply to other contexts. Credibility asks whether the findings are believable.

Lincoln and Guba proposed these criteria. They are asked many times, so learn the four pairs.

Triangulation

Triangulation looks at the question from different angles.

• Different researchers, different methods or different data.
• Gives a more complete picture.
• Builds confidence in the findings.

When different perspectives agree, the researcher can trust the finding more. Triangulation is expected to produce knowledge that is more complete and sound.

Documents, history and other methods

Historical research

Historical research collects facts about the past from records.

• Sources: written records, diaries, letters, newspapers, relics, oral testimony.
• Cannot be repeated or controlled.
• Limitation: universal generalisation is not possible.
• Poor analysis: over-generalisation, over-simplification, ignoring social context.

A study of the lifestyle of Indian people between 1900 and 1947 is historical research.

Personal observation and focus groups are not sources of data in historical research.

Primary, secondary and tertiary sources

Sources of data are of three kinds, by how close they are to the original.

• Primary: original or first-hand, such as films, posters, gazettes, interviews.
• Secondary: interprets primary sources, such as newspapers and magazines.
• Tertiary: derived from both primary and secondary sources.
• Mailed questionnaire, observation and interview give primary data.

Data collected first-hand by the researcher from respondents is primary data. Using what others have already collected is using a secondary source.

Content analysis

Content analysis studies texts and documents, usually by counting categories.

• Classically a quantitative approach.
• Transparent, unobtrusive and flexible.
• Coding is crucial.
• Allows some longitudinal analysis.
• Replication is feasible.

Coding means sorting content into categories. The coding rules are clear, so others can check them.

It is unobtrusive because the researcher does not disturb the people who produced the text.

Action research report and non-scientific ways

An action research report follows a set order. Some ways of knowing are not scientific.

• Report: introduction, objectives, methodology.
• Then intervention and its impact, results, interpretation, suggestions.
• Not scientific: tradition, personal experience, intuition, authority.
• Scientific method is empirical, objective, systematic and predictive.

The teacher acts on the problem before there are results, so the intervention comes before the results in the report.

Tradition, personal experience, intuition or authority are subjective ways of knowing. They are not methods of scientific research.

How a study is planned, the kinds of variables, the four scales, and reliability and validity.

Research designs

What a research design is

A research design is the overall plan or blueprint of a study.

• Conceptual framework in which research is conducted.
• Provides a blueprint.
• Sets the boundaries of the research activity.
• Lets the investigator anticipate problems.
• Components: comparison, control, manipulation, generalisation.

A design tells what will be studied, how, with whom and how the data will be analysed. A synopsis is only a summary of the proposal.

The time frame of a study is also called its reference period.

Pilot study

A pilot study is a small trial run before the main study.

• Also called a feasibility study.
• Tests the tools and the plan.
• Finds problems early.
• In a survey, it comes after designing the questionnaire.

The survey stages are: design the questionnaire, pilot test it, distribute it, analyse the responses.

Cross-sectional and longitudinal designs

A cross-sectional study looks at one point in time. A longitudinal study follows change over time.

• Cross-sectional: a snapshot, such as a survey at one time or a census.
• Longitudinal types: trend, cohort, panel and prospective.
• Cross-sectional is not a type of longitudinal study.
• Content analysis of documents from different periods is longitudinal.

A population census collects data from everyone at one point in time. So it is an example of a cross-sectional study.

Longitudinal studies meet the study population several times at regular intervals. The gaps can be a week or more than a year. So there is a need to meet more than once.

Trend, cohort and panel studies

These three longitudinal studies differ in who is studied at each round.

• Trend: different samples from a general population.
• Cohort: different samples from a specific, well-defined population.
• Panel: the same people each time.
• Prospective: follow people forward in time.

A cohort is a group that shares a specific feature, such as the same year of birth. Do not swap the descriptions of cohort and trend. Questions often do.

Longitudinal, panel and cohort studies can suffer from sample attrition, since people drop out over time. Cross-sectional studies do not have this problem.

Experimental and quasi-experimental designs

Experimental designs manipulate and control. Quasi-experimental designs lack full random assignment.

• Experimental: manipulation and control of the independent variable.
• Quasi: non-equivalent control group design.
• Quasi: interrupted time-series designs.
• Solomon four group and post-test only designs are true experiments.

In a quasi-experiment the researcher cannot randomly assign people to groups. The groups already exist, such as two classes.

The group that experiences the manipulated independent variable is the experimental group. The group that does not is the control group.

Major designs and design terms

Social research uses a few major designs, and some terms are not designs at all.

• Major designs: experimental, cross-sectional, longitudinal, case study, comparative.
• Comparative design: two contrasting cases, identical methods.
• Not designs: prescriptive and manipulative.
• Particularistic research: one place, one time, one question.

A case study looks closely at one case. A comparative design places two or more contrasting cases side by side.

In the modelling approach, the problem is represented by a mathematical model, and the solutions come from that model.

Mixed method designs

Mixed method designs use both qualitative and quantitative methods.

• QUAN then qual: explanatory sequential.
• QUAL then quan: exploratory sequential.
• Nested (one inside the other): embedded design.
• Both together: convergent design.

Capital letters show the main method. An arrow shows the order. A plus sign means both at the same time.

In the explanatory sequential design, numbers come first and then interviews explain them.

Blind and double-blind designs

Blinding hides group identity to prevent bias.

• Single blind: either the investigator or the participant does not know the treatment.
• Double blind: neither knows.
• Placebo group gets a fake treatment.
• Experimenter bias is prevented by blinding.

The word 'either' in a question points to single blind. The words 'neither' or 'both' point to double blind.

Variables

Independent and dependent variables

The independent variable is the cause. The dependent variable is the effect.

• Independent: manipulated, causes a change.
• Dependent: the outcome that is measured.
• Statistical analysis uses the dependent variable's observations.
• An independent variable can have any number of levels.

In 'effect of teaching methods on achievement', teaching method is the independent variable and achievement is the dependent variable.

An independent variable can be qualitative or quantitative. It does not have to have only two levels.

Intervening, extraneous and control variables

Some variables sit between cause and effect, or disturb the result.

• Intervening variable: comes between cause and effect.
• Extraneous variable: an unwanted outside influence.
• Control variable: kept constant on purpose.
• Extraneous variables are not always controlled.

In a study of two yoga exercises and body fat, food habit is an intervening variable. It can change fat even though it is not the exercise.

Intervening variables can be controlled by statistical, physical and selective manipulation.

Dichotomous, discrete and continuous variables

Variables differ in the kind of values they take.

• Dichotomous: only two categories, such as yes or no.
• Discrete: separate whole-number values, such as number of children.
• Continuous: any value in a range, such as height and weight.
• A dependent variable can be dichotomous. So can an independent variable.

Number of heads in ten tosses is discrete. Weight and height are continuous. Marks in a test with whole-number marks are discrete.

Dichotomous means two categories. A nominal variable can have more than two categories.

Measurement

The four scales of measurement

The four scales go from the simplest to the most advanced: nominal, ordinal, interval, ratio.

• Nominal: names or categories only.
• Ordinal: ranks, order but unequal gaps.
• Interval: equal gaps, no true zero.
• Ratio: equal gaps and a true zero.

Ratio is the highest scale. It allows all operations, including ratios of values. Nominal is the lowest.

Ranking or grading belongs to the ordinal scale. A teacher who gives ranks on the basis of scores is using it.

Examples of each scale

Learn one or two examples of each scale. They come again and again.

• Nominal: gender, religion, caste, blood group, type of forest.
• Ordinal: rank of an army officer, level of irritation, satisfaction level.
• Interval: temperature in Celsius or Fahrenheit, date of birth, time of day.
• Ratio: weight, height, distance, crop yield, age.

Religion and caste are categories only, so they are nominal. They cannot be ranked.

Interval scales have no absolute zero. 00:00 does not mean 'no time', and year 0 does not mean 'no year'.

True zero and the order of scales

Only the ratio scale has an absolute zero.

• Nominal and ordinal: no true zero.
• Interval: zero is arbitrary.
• Ratio: zero means none of the quantity.
• Nominal data cannot be ranked.

A statement that says zero is arbitrary in a ratio scale is wrong. A statement that says the nominal scale allows ordering is also wrong.

Attitude scales

Attitude scales turn opinions into numbers.

• Likert scale: summated rating scale, the easiest and most common.
• Thurstone: a panel judges the items (consensus approach).
• Guttman: cumulative scale.
• Factor scales: rely on item intercorrelation.

In a Likert scale the respondent says how much they agree with each statement. The scores are added.

Steps in building a tool: construct items, select items for the primary draft, item analysis, select items for the final draft, then norms.

Reliability and validity

Reliability

Reliability means that a measure gives consistent results.

• Test-retest: same test repeated over time.
• Inter-rater: different people score the same thing.
• Parallel forms: two versions of the test.
• Split-half: one test divided into two halves.

Stability of results across time shows reliability. Agreement between observers shows inter-observer reliability.

A reliable test can still be wrong. Reliability does not ensure validity, but a valid test must be reliable.

Measurement validity and construct validity

A measure is valid when it measures what it claims to measure.

• Measurement validity: does the measure reflect the concept.
• Construct validity: does it truly represent the idea, such as intelligence.
• Convergent validity: different methods give similar results.
• Validity = accuracy.

Two researchers who use different methods and get similar results show convergent validity.

Internal, external and ecological validity

These three validities answer three different questions.

• Internal validity: was it really the cause that produced the effect.
• External validity: can results be generalised to other people and settings.
• Ecological validity: do findings apply to everyday natural settings.
• Internal validity relates to causality.

When you want to generalise results to a different population, your study needs external validity.

A finding from an artificial laboratory may not hold in real life, so ecological validity may be weak.

Threats to internal validity

Threats are other things that could explain the result instead of the treatment.

• History: outside events during the study.
• Maturation: natural change such as boredom and fatigue.
• Selection: groups differ from the start.
• Attrition or mortality: participants drop out.
• Testing and instrumentation.

Testing means that taking a test earlier affects the later score. Instrumentation means that the measuring tool changes during the study.

Randomisation and generalisability are not threats to internal validity. Placebo effect and Hawthorne effect are separate ideas.

Researcher and respondent biases

Bias creeps in from the researcher and from the people studied.

• Experimenter bias: researcher subtly affects participants.
• Biosocial effect: researcher's age, gender or caste affects behaviour.
• Response bias: predictable answering, whatever the question.
• Social desirability: impression management, self-deception positivity.
• Nondifferentiation: same answer to every item.

Impression management means deliberately showing a good image. Self-deception positivity means honestly but wrongly believing a good view of oneself.

How a sample is chosen, how survey questions work, and the tools for collecting data.

Sampling

Population, sample and sampling terms

Sampling is the process of selecting a small number of elements from a larger group.

• Population: the larger group under study.
• Sample: the small group chosen from it.
• Sampling frame: the list of population elements available for selection.
• Sample unit: a single element of the sample.
• Sampling error: the gap between the sample and the population.

A sample is useful only if it represents the population. A good frame is the starting point for choosing it.

Probability sampling

In probability sampling every unit has a known chance of being chosen, so chance and not the researcher decides.

• Simple random sampling: equal and independent chance.
• Systematic sampling: select every i-th item from a list.
• Stratified random sampling: divide into strata, then sample each.
• Cluster sampling: divide into groups, then pick whole groups.

The fishbowl draw is a method of drawing a random sample. All names go in a bowl and are drawn one by one.

Probability sampling is mostly used in survey, experimental and correlational research. It lets the results be generalised.

Non-probability sampling

In non-probability sampling the researcher or the situation decides who is chosen.

• Convenience: easiest to reach, also called accidental or haphazard.
• Purposive: chosen by judgement, also called judgemental or deliberate.
• Quota: fixed numbers of each type, like non-random stratified sampling.
• Snowball: participants recommend other participants.
• Dimensional sampling is also non-probability.

Non-probability samples use available respondents. They do not accurately reflect the population, so results cannot be generalised in the same way.

Key informant sampling gets information from people who know the population well. It is also a non-probability method.

Choosing the right sampling method

The kind of research and the kind of population decide the method.

• Heterogeneous population: stratified random sampling.
• Empirico-inductive (qualitative) research: non-probability sampling.
• Least bias: simple random sampling.
• Most bias: snowball sampling.
• Hidden populations: snowball and key informant sampling.

A heterogeneous population has different groups. Simple random sampling may miss some groups, while stratified sampling gives each group a place.

A big sample does not remove the bias of a wrong selection procedure. It only repeats the same bias on a larger scale.

Sampling problems in online surveys

Online surveys grow fast, but their samples are often biased.

• Internet users are not always an unbiased sample.
• Many people use more than one internet service provider.
• A household may have one computer but several users.
• Convenience samples are simply available people.
• Online tools make surveying and analysis easier.

Some groups have no internet access, so they are left out. This is selection bias.

Survey research

Closed-ended and open-ended questions

A closed-ended question gives fixed options. An open-ended question lets the person answer freely.

• Closed: easy to process, easy to complete, less variation in recording.
• Closed: loss of spontaneity, hard to list all answers.
• Open: respondents answer in their own terms.
• Open: good for exploring new areas, allow unusual answers.
• Open: time-consuming, answers are subjective.

A questionnaire with no scope for the respondent's own view is closed-ended. Closed questions cannot allow unusual answers.

Writing clear survey options

Answer options must have clear meaning, so every person reads them the same way.

• 'Very often', 'quite often', 'not very often' are vague.
• 'Not at all' is clear.
• Options must be exhaustive and not overlap.
• Do not suggest an answer in the question.

Words like 'often' mean different things to different people, so they are ambiguous. A number, such as 'twice a week', is clear.

Survey errors, stages and modes

A survey has stages, and error can enter at several points.

• Stages: design the questionnaire, pilot test, distribute, analyse.
• Errors: sample error, data collection error, data processing error.
• Telephone survey: reduces the cost of data collection.
• A schedule is a form filled in by the investigator.

Sample error comes from observing a sample and not the whole population. Data collection error comes from wrong collection. Processing error comes from wrong coding or entry.

Tools and data

Interview and questionnaire

A questionnaire is a written data collection instrument. An interview is a spoken one.

• Structured interview: questions prepared beforehand.
• Introductory questions should relate to the research topic.
• General questions come before specific ones.
• Error: wrong recording, influencing responses, poor environment.

Error creeps into interview data when the interviewer fails to record answers accurately, influences the respondent or does not set up a proper interview environment.

Online focus groups have limits too. Drop-outs can spoil the data quality.

Personal involvement of the researcher

Methods of data collection differ in how close the researcher is to the people.

• Highest: participant observation.
• Then observation.
• Then unstructured interview.
• Then structured interview.
• Lowest: questionnaire.

In participant observation the researcher joins the group. In a questionnaire the researcher is absent, and the respondent fills it alone.

Tests used in research

Tests are tools that measure ability or personality.

• Maximum performance tests: achievement, intelligence and aptitude tests.
• Typical performance: personality, projective and attitude tests.
• Maximum performance tests have right and wrong answers.
• Tool construction: items, primary draft, item analysis, final draft, norms.

In a maximum performance test the person is asked to do their best. In a typical performance test the person is asked what they usually do or feel.

Handling the data

Raw data must be prepared before analysis.

• Recorded audio or video is turned into text by transcription.
• Quantitative analysis can use Excel, Access and SPSS.
• Word and PowerPoint are not analysis tools.
• Primary data: collected first-hand.

Data recorded by audio or video cannot be analysed directly. It must be written out first. This step is transcription.

Describing data, correlation, the normal curve, hypothesis tests, p-values and errors.

Describing data

Central tendency and dispersion

Central tendency shows the typical value. Dispersion shows how spread out the values are.

• Central tendency: mean, median, mode.
• Dispersion: range, quartile deviation, standard deviation.
• Second decile is a measure of position.
• The sum of deviations about the mean is always zero.
• Standard deviation is the most sophisticated measure of dispersion.

With a few extreme scores the mean is pulled away, so the median is the best measure. The mode is not the best choice in that case.

If a constant is added to every score, the mean goes up by that constant, but the standard deviation stays the same. Adding 2 to a mean of 45 gives 47, and the standard deviation stays 10.

Position measures: quartiles, deciles and percentiles

These measures show the position of a value in ordered data.

• Median = 50th percentile = 2nd quartile.
• 6th decile = 60th percentile.
• 3rd quartile = 75th percentile.
• Higher percentile means a higher value.

So the increasing order of median, 6th decile, 70th percentile and 3rd quartile is: median, 6th decile, 70th percentile, 3rd quartile.

Skewness

Skewness shows which side the long tail of a distribution is on.

• Positive skew: tail on the right, mean > median > mode.
• Negative skew: tail on the left, mean < median < mode.
• The mean is always pulled toward the tail.

In a positively skewed distribution, a few very high scores drag the mean up. So the mean is the largest of the three.

Formulas that the questions use

Many questions only need one short formula. Learn these.

• Coefficient of variation = (standard deviation / mean) x 100.
• Variance = mean of squares minus square of the mean.
• Standard error = sigma divided by the square root of n.
• Binomial: mean = np, variance = npq.
• Covariance = r x S(X) x S(Y).

In the coefficient of variation, if variance is given, take its square root first to get the standard deviation.

For a binomial, q = variance divided by mean. Then p = 1 - q and n = mean divided by p.

The normal curve

The normal curve is a bell shape. The standard normal variable z has mean 0 and standard deviation 1.

• Between -1 and +1: about 0.68.
• Left of 0: 0.5.
• From 0 to +1: about 0.34.
• From +1 to +2: about 0.14.
• Beyond +3: very small, about 0.001.

The chance of a value falling more than 3 standard deviations from the mean is very small in normally distributed data.

The area falls quickly as you move away from the middle. Remember 0.34, 0.14, 0.02 and 0.001, and ordering questions become easy.

Graphs and charts

Each graph suits a particular kind of data.

• Histogram: frequency distribution, bars touch.
• Line chart: time-series data, variations over time.
• Box plot: median, quartiles and outliers.
• Bar chart: categories.
• Scatter plot: relation of two variables.

A box plot cannot show correlation. It shows the median as a line in the box and outliers as points beyond the whiskers.

Relationship between variables

Correlation

Correlation shows how strongly two variables move together.

• The main aim is to find the association among variables.
• Value runs from -1 to +1.
• The sign shows direction. The size shows strength.
• Positive: when one rises, the other rises.

Correlation shows association only. It does not prove that one variable causes the other.

Pearson, Spearman and Kendall

The right correlation depends on the scale of the data.

• Pearson's r: interval or ratio data, linear relation.
• Spearman's rho: ranks, ordinal data.
• Kendall's tau: rank correlation.
• Pearson's r is not suited to nominal variables or curved relations.

To calculate Spearman's rho you enter ranks. If you have actual scores, first change them into ranks.

Rank correlation formula: rho = 1 - 6 x (sum of d squared) / (n x (n squared - 1)). Here d is the difference in ranks.

Correlation, causation and partial correlation

To claim cause, you must rule out other possible causes. This is the internal validity rule.

• Cause needs co-variance, time order and no other cause.
• Partial correlation controls other variables.
• First-order partial correlation: three variables.
• Second-order partial correlation: four variables.

The number of variables needed grows as you control more. A mean needs one variable, a correlation coefficient two, a first-order correlation three and a second-order correlation four.

Hypothesis testing

Null and alternative hypotheses

The null hypothesis says there is no effect or no difference. The alternative says there is one.

• Null hypothesis: written H0.
• Alternative hypothesis: written H1 or Ha.
• Both are statements about population parameters.
• Alternative can be: not equal, greater than, or less than.
• A hypothesis must be falsifiable.

If the evidence is strong enough, the researcher rejects H0 and accepts the research hypothesis. If it is not, the research hypothesis is not supported.

The null is only assumed true until the data show otherwise. The statement that the alternative says 'no relationship' is wrong.

One-tailed and two-tailed tests

The form of the alternative hypothesis decides which tail of the curve is used.

• Ha: mean not equal to a value: two-tailed.
• Ha: mean greater than a value: right-tailed.
• Ha: mean less than a value: left-tailed.
• 'Is the mean lower than...' means a left-tailed test.

If the researcher wants to test whether the population mean is lower than a stated value, the rejection region is in the left tail.

p-value and level of significance

The p-value is the chance of results this extreme if the null hypothesis is true.

• Small p-value: evidence against the null.
• Significant at 5%: p less than 0.05.
• Significant at 1%: p less than 0.01.
• Smaller p means a higher level of significance.
• p-value depends on the sample size.

A p-value of 0.43 or 0.27 is large, so the null looks plausible. A p-value of 0.005 is below 0.01, so it is significant even at the 1% level.

A very small sample may not show an important difference as significant. A larger sample gives a smaller p for the same difference.

Type I error

A Type I error means rejecting a true null hypothesis. It is a false alarm.

• Probability of Type I error = significance level (alpha).
• A larger alpha means a larger risk of this error.
• Risk is higher at 0.05 than at 0.01.
• Highest risk: alpha 0.10. Lowest: alpha 0.01.

If you set a strict level such as 0.01, you reject the null less often when it is true. So the risk of a Type I error falls.

Which test to use

The data and the number of groups decide the test.

• Two group means: t-test.
• Three or more group means: analysis of variance (ANOVA), F-test.
• Counts or frequencies: chi-square.
• t-test also tests significance of a correlation or regression coefficient.
• F uses numerator and denominator degrees of freedom.

The F statistic is the ratio of between-group variance to within-group variance. That is why it needs two kinds of degrees of freedom.

Chi-square compares observed counts with expected counts. It also tests the independence of attributes.

Parametric and non-parametric tests

Parametric tests assume a distribution. Non-parametric tests do not.

• Parametric: t-test, F-test, Z-test, Pearson's r.
• Non-parametric: Mann-Whitney U, Kruskal-Wallis H, Wilcoxon, Kendall's W, rank correlation, chi-square.
• Parametric conditions: normal data, interval or ratio scale, equal variances.
• Non-parametric tests are distribution-free and quicker.
• Bartlett's test checks homogeneity of variances.

A sample size above 30 is only a rule of thumb. It is not a condition for a parametric test.

The Mann-Whitney test is the U test and the Kruskal-Wallis test is the H test. Both are non-parametric.

Thesis and article writing, referencing styles, ICT tools, research metrics and research bodies.

Writing a thesis, article and paper

Formats of presentation

Research is shared in different formats, and each has its own feature.

• Thesis or dissertation: systematic, prescribed format.
• Research paper: summary of the research done.
• Seminar: reflective deliberation on a theme.
• Workshop: target-oriented, group-based work.
• Position paper: highlights issues and depicts the present status.

Target-related specifications are necessary in workshops, because a workshop works towards a skill or task.

Poster sessions at conferences give better chances for inter-personal interaction. Visitors can talk to the author one to one.

Parts of a thesis and a research report

A thesis has fixed parts, and the supervisor certifies that it is original.

• Prefatory parts, title page and table of contents.
• Terminal chapters and references.
• Supervisor's certificate ensures originality of the work.
• Chapter scheme is approved by the research degree committee.
• Financial accounts and utilisation certificates are not part of the report.

A dissertation begins with the acknowledgement, then the literature review, then the research methods, then the discussion.

The style of writing a thesis or article should be scientific in language. It should be precise and objective, and not decorative.

Sections of a research paper

A research paper follows a fixed order, from abstract to conclusion.

• Abstract.
• Introduction: gives the theoretical reasons for the study.
• Procedures (methods).
• Results.
• Conclusions.

A research abstract is a must in a research article, a seminar paper and a thesis. A synopsis is itself a short summary, so it needs no separate abstract.

To know why researchers did a study, read the introduction of their article.

Research proposal and a good report

A proposal says what you will do. A good report shows originality and logic.

• Proposal: statement of the problem, objectives, review of literature, methodology.
• Data fabrication is never part of a proposal.
• A good report shows originality and solves an intellectual problem.
• It gives a logical analysis, and may include charts and tables.

Making up data is misconduct. It can never be a stage of a proposal.

A report should use clear language, not jargon that adds complexity.

Reading skills and study skills

Reading for research follows the SQ3R order.

• Survey, Question, Read, Recall, Review.
• Elaborative interrogation: asking why a fact is true.
• Practice testing: testing yourself.
• Self-explanation: explaining a section to yourself.
• Interleaved practice: mixing different kinds of problems.

SQ3R means: survey the whole text first, make questions, read, recall what you read, and then review.

Peer review and predatory journals

Good journals check work with peer review. Predatory journals do not.

• Blind review: the reviewer does not know the author.
• Predatory journals lack genuine peer review.
• They use aggressive solicitation and promise rapid publication.
• A clear retraction policy marks a genuine journal.
• They harm the reputation and career of the researcher.

Predatory journals put their own profit ahead of scholarship. They give misleading information and lack transparency.

Referencing

Why we reference and the main styles

References credit the sources and let the reader check the work.

• Reference lends authenticity to the content.
• Real styles: APA, MLA, Chicago, Harvard, Vancouver, IEEE.
• Not real styles: Cambridge, New York and Roman.
• APA is seen as a variant of Harvard, an author-date style.
• IEEE uses numbers in square brackets.

In-text citations differ. APA gives the surname and year, such as (Sharma, 2006). MLA names the author and gives page numbers.

APA order of a reference

In APA style the author's last name comes first, then the year.

• Journal article: author, year, title, journal, volume (issue), pages, DOI.
• Book: author, year, title, publisher.
• Conference: author, year, name of conference, location.
• DOI means Digital Object Identifier, written at the end.

The first name is not written first. Only the initials of the first name follow the last name. The year does not come after the page number.

The journal name is written in italics in a journal article reference.

Abbreviations used in references

These Latin short forms appear in footnotes and references.

• et al.: and others.
• ibid.: in the same place, same work cited just above.
• op. cit.: in the work cited.
• id. or idem: the same.
• cf.: compare.
• vid. or vide: see, refer to.
• viz.: namely.
• passim: here and there.
• ante: before. post: after.
• fn.: footnote.

Learn them in pairs. Ante and post are opposites. Ibid. and op. cit. both point to a source already given.

ICT, metrics and research bodies

ICT tools in research

ICT tools help to find, manage, analyse and check research.

• WAN gives wider coverage and access to information.
• Bibliographic software: EndNote, Reference Manager, Mendeley.
• Turnitin is a plagiarism-detection tool.
• Analysis software: Excel, Access, SPSS.
• PubMed is a database for biomedical and life sciences.

When choosing ICT for research, the tool should be appropriate, affordable and available in the department.

The internet is the largest WAN. A LAN covers only a small area, such as one building.

Research metrics: h-index and others

The h-index measures both how many papers and how often they are cited.

• h-index: the largest h such that h papers have at least h citations each.
• Impact factor: a journal's influence and prestige.
• Citation count: impact of an article or author.
• i10-index: number of papers with at least 10 citations.

Example: citations 21, 12, 4, 2, 1. The third paper has 4 citations, which is at least 3. The fourth paper has 2, which is less than 4. So h = 3.

For a method: sort citations from high to low. Find the last rank whose paper has at least that many citations.

Research bodies and funding

Many public bodies support research in India.

• Public funders: ICAR, ICHR, ICMR.
• Tata Trust and Population Foundation of India are not public-sector agencies.
• STRIDE was initiated by UGC.
• Order of setting up: IIAS 1964, ICSSR 1969, ICHR 1972, ICPR 1977, NCRI 1995.

STRIDE stands for Scheme for Trans-disciplinary Research for India's Developing Economy.

Public-sector agencies are run or funded by the government.

Meta-analysis and research forms

Meta-analysis combines the results of many studies.

• Meta-analysis: combines studies.
• File drawer problem: unpublished studies are missing.
• Triangulation: convergence of findings from multiple methods.
• Qualitative: interpretive. Quantitative: statistical.

The file drawer problem arises because studies with no clear result often stay unpublished in a drawer. A meta-analysis then sees only part of the evidence.

Honest conduct in research, rules for human participants, and the UGC 2018 rules on plagiarism.

Ethical conduct in research

What research ethics is

Research ethics is a set of dos and don'ts that protect participants and keep research honest.

• Protects the rights and welfare of participants.
• Gives rules for ethical conduct.
• It is not a guide to write a flawless report.
• It is not the same as research methodology.

Ethics asks whether the research is done in the right way. Method asks whether it is done in a way that works.

Stages where ethics matters

Ethics matters at every stage, but the keyed answers point to the stages that deal with people and results.

• Data collection.
• Data analysis and procedures used.
• Reporting of findings.
• Hypothesis testing and interpretation.

These are the stages where the researcher meets participants, handles data and tells the result. This is where cheating or harm is most likely.

Choosing the problem, defining the population and choosing the design are earlier planning steps.

Rules for research with people

A study with human participants must protect them.

• Informed consent: people join knowing what the study involves.
• No physical or psychological harm.
• No coercion and no excessive inducement.
• Confidentiality of information.
• No invasion of privacy.

Sharing confidential information about subjects with others is a breach of ethics.

Excessive inducements pressure people to join, so they are an ethical issue. One NTA key also treated offering incentives as unethical.

No racial comments, informed consent and privacy are ethical concerns. Sample size and the researcher's economic status are not.

Deception and debriefing

Deception means misleading participants. It must be corrected afterwards.

• Active deception: telling something false to mislead.
• Passive deception: withholding information.
• Dehoaxing: telling participants afterwards about the deception.
• Debriefing explains the study to participants after it.

Dehoaxing aims to reduce the harm that deception may cause. Debriefing is a good practice and not an ethical problem.

In ethnography, a rigid use of ethical rules is not possible. Informed consent can become theoretical, since telling everything beforehand may change behaviour.

Fabrication, falsification and plagiarism

These are the three serious forms of research misconduct.

• Fabrication: making up data.
• Falsification: changing or misreporting real data.
• Plagiarism: using someone else's work as your own.
• Filling many forms yourself without real data collection is fabrication.

Plagiarism can be avoided by proper citation, by paraphrasing correctly with citation, and by using a plagiarism checker.

Crediting only the main author, or replacing the original author's name, is dishonest.

Error of omission, technical lapse and unethical act

Not every mistake in a thesis is unethical.

• Leaving out the reason for a statistical technique: error of omission.
• Wrong format or referencing style: technical lapse.
• Not acknowledging respondents' help: unethical act.
• Acknowledging the supervisor: part of ethicality.

An error of omission means leaving out something that should be there. An error of commission means doing something wrong, such as fabricating data.

A seminar paper with very old references shows that the teacher has not updated the sources. It is not an ethical lapse.

Researcher effects and codes of ethics

Who the researcher is, and how they act, can affect participants.

• Biosocial effect: sex, age, caste or race of the researcher.
• Psychosocial effect: the researcher's attitude.
• Hawthorne effect: reaction to special attention.
• APA: American Psychological Association.
• British Sociological Association and Social Research Association have useful codes.

When the researcher's sex affects participants' behaviour, it is a biosocial effect. When the researcher's attitude does, it is a psychosocial effect.

Fair use of copyrighted material

Teaching allows only a small part of copyrighted work to be used.

• For a MOOC video: up to 10% or 3 minutes, whichever is less.
• Always cite the source.
• Using more needs permission.

This is a fact to remember as a number. It applies to developing a video for a MOOC with a copyrighted video.

Plagiarism and the UGC Regulations 2018

The four levels of plagiarism

UGC Regulations 2018 set four levels of plagiarism by the percentage of similarity.

• Level 0: up to 10%.
• Level 1: above 10% to 40%.
• Level 2: above 40% to 60%.
• Level 3: above 60%.

14% is Level 1. 42% and 57% are Level 2. 63% and 71% are Level 3. 38% is Level 1.

Check the boundaries. 40% is the top of Level 1 and 60% is the top of Level 2.

Penalties for students

The penalty for a student grows with the level.

• Level 0: no penalty.
• Level 1: submit a revised script within 6 months.
• Level 2: not allowed to submit a revised version for one year.
• Level 3: registration for that programme is cancelled.

The student must give an undertaking with the thesis that it is original work, free of plagiarism. The supervisor must also certify that the work is plagiarism free.

Penalties for faculty and researchers

Faculty members face heavier penalties as the level rises.

• Level 1: withdraw the manuscript.
• Level 2: withdraw the manuscript and lose one annual increment.
• Level 2: also not allowed to supervise new scholars for two years.
• Level 3: withdraw, lose two successive annual increments, no supervision for three years.

'Denial of one annual increment' points to Level 2. Remember that Level 3 is two increments.

What is left out of the similarity check

Some parts do not count when similarity is measured.

• Common knowledge or coincidental terms up to 14 consecutive words.
• Generic terms, laws and standard equations.
• Bibliography and references.
• Abstract, observations and conclusions are checked.

The number 14 is a fact that is asked directly. Remember '14 consecutive words'.

Who handles plagiarism

Institutions set up panels to handle plagiarism cases.

• Institutional Academic Integrity Panel (IAIP): four members.
• Departmental Academic Integrity Panel: headed by the Head of the Department.
• Proceedings can start from an examiner's findings.
• The institution can take suo moto notice (on its own).

A higher education institution does not wait for a complaint. It may act on an examiner's finding or take notice by itself.