2.4 Research methodology

Data representation turns observations into a table, chart or graph so that a reader can see values, comparisons and patterns. The choice depends on whether the variables are categorical, discrete, continuous or ordered.

Label axes, units and categories, and preserve a scale that does not exaggerate differences. A display summarises evidence; it does not create a causal relationship or repair biased sampling.

A bar chart can compare counts in separate categories, while a histogram groups continuous values into intervals. A line graph is useful for change across an ordered variable, but joining unrelated categories implies a false sequence.

A polished graph is not automatically valid. Check what each mark represents, whether the scale is honest, and whether the display hides missing data or unequal group sizes.

The mean, median and mode describe different centres: the arithmetic average, the middle ordered value and the most frequent value. Range and standard deviation describe spread, not the size of the centre.

Use the median when extreme values would pull the mean away from a typical case. A small spread means scores cluster around the chosen centre; it does not by itself prove a measure is reliable or valid.

For 2, 3, 3, 4 and 18, the mean is 6 and the median is 3. The outlier is part of the evidence: report it, check how it arose, and justify any decision to analyse it separately rather than deleting it silently.

There is no universally best summary. Match the statistic to the scale, distribution and research question; never treat a single central value as the whole dataset.

For ordered data, the semi-interquartile range is (Q3 − Q1) / 2, so it describes the spread of the middle 50% and resists extreme scores. Standard deviation describes typical distance from the mean and is most informative alongside the distribution and mean. For 2, 3, 3, 4, 18, Q1 = 2.5 and Q3 = 11, so SIQR = (11 − 2.5) / 2 = 4.25; state the units of the original scores when reporting any measure of spread.

An inferential test compares an observed result with what would be expected if the null hypothesis were true. A p-value is the probability of a result at least this extreme under that model, not the probability that the null is true.

At a chosen significance level such as 0.05, a result below the threshold is evidence against the null; it does not prove a theory or measure the size of an effect. A critical-value decision must match the test and tail.

A Type I error is a false positive: rejecting a true null. A Type II error is a false negative: failing to reject a false null. Tightening the threshold can reduce Type I risk while making Type II errors more likely.

‘Not significant’ is not proof of no effect, and ‘significant’ is not proof of importance. Report the decision, uncertainty and design limits together.

Choose the test from design and data: an unrelated t-test compares two independent means when parametric assumptions are suitable; a related t-test compares paired means; Mann–Whitney and Wilcoxon are corresponding rank-based alternatives; chi-square tests association between categorical variables; a correlation coefficient describes direction and strength of a relationship. Statistical significance addresses compatibility with the null model, while effect size addresses magnitude. Neither repairs bias, confounding or a non-causal design.

Thematic analysis organises qualitative material—such as interviews, diaries or messages—into codes, categories and broader themes. In an inductive analysis, themes are developed from the data rather than imposed in advance.

The researcher moves from repeated features to an interpretation of what they mean for the question. Reflexivity matters because prior expectations can influence which passages are noticed, grouped or treated as exceptions.

A set of interviews might contain repeated references to blame, fear and control. Those codes can form a theme about coercion, but the analyst should show how excerpts support it and whether disconfirming accounts were considered.

A theme is not simply a topic label or a frequency count. It is an evidence-supported pattern with a clear coding route, and qualitative richness does not remove the need for transparency.

A defensible sequence is: become familiar with the full dataset; code relevant features; group related codes into candidate themes; review themes against coded extracts and the dataset; define and name each theme; then write the analysis using extracts and interpretation linked to the research aim. Researchers should record revisions and actively look for contradictory material rather than forcing every response into the first pattern noticed.

Credibility is the trustworthiness of a qualitative account: whether its interpretation is believable given the data, method and researcher. Transferability asks whether the richly described insight can inform another context, not whether it statistically represents a population.

Transparency about sampling, procedure, analysis and limitations lets a reader judge fit. Triangulation can compare datasets or researchers, while reflexivity makes the researcher's influence visible rather than pretending interpretation is neutral.

An interview study of refugee experiences may be credible when the coding process and participant context are explicit, yet not transferable to every migration setting. The boundary is about fit and detail, not a simple pass/fail label.

Credibility is not the same as replicability, and transferability is not automatic generalisability. State which claim is being made and which evidence supports it.

Research methods are the planned ways psychologists collect and analyse evidence, such as experiments, interviews, observations, surveys and case studies. The method must operationalise the variables and fit the kind of claim being made.

Experiments can support causal inference when manipulation, comparison and control are credible. Interviews and observations can reveal meaning and behaviour in context, but researcher influence, sampling and interpretation become central validity questions.

To study whether sleep changes reaction time, an experiment can manipulate sleep opportunity and measure a defined response. To understand how students experience sleep loss, a semi-structured interview may be more informative but cannot alone establish causation.

No method is automatically superior. Identify the research question, the evidence needed, and the trade-off between control, ecological validity, ethics and transferability before judging the design.

Different methods answer different questions: experiments manipulate an independent variable for causal inference; observations record behaviour in a setting; surveys collect standardized self-report; interviews explore accounts in depth; correlational studies measure relationships without manipulation; case studies integrate multiple sources around one bounded case. Selection must fit the claim and access conditions, while ethics, demand characteristics, researcher effects, confounds and external variables limit what can be concluded.

A sample is selected from a target population so that a study can collect evidence within practical limits. Opportunity and volunteer samples are convenient; random and stratified methods aim for better representation; snowball sampling reaches hidden populations.

Choose the method from the research aim and access conditions. A random draw can still be unbalanced by chance, while a stratified design matches selected population proportions only for characteristics that were measured and used.

A study of an uncommon group may begin with one trusted participant and use referrals, gaining access but narrowing the network. A campus volunteer advert is faster but can overrepresent people who are willing and interested in the topic.

A large sample is not automatically representative. State who could be selected, who was actually selected, and how that route limits generalisation and introduces bias.