Practical investigation — Edexcel A-Level Psychology

The cognitive practical is a laboratory experiment you design, run and write up yourself, and the marks follow the decisions rather than the result. Settle the design first, because it fixes the test: two separate groups of participants point to Mann-Whitney U, while the same people tested twice point to the Wilcoxon signed ranks test, and both rank scores instead of assuming a normal distribution. Operationalise the variables so another student could repeat you, counterbalance to hold order effects off, standardise instructions so experimenter effects do not creep in, and disguise the aim enough that demand characteristics do not write your results for you. Summarise with a mean or median alongside range or standard deviation, graph the pattern, then compare the observed value with the critical value before you claim anything.

A correlation measures two co-variables in the same people and asks whether they move together, so nothing is manipulated and nobody is allocated to a condition, and the write-up must never claim that one thing caused the other. Rank each co-variable separately, subtract the paired ranks, square the differences and total them before applying the formula, then read the coefficient twice, once for direction from its sign and once for strength from its size, and check it against the critical value for the number of pairs at the five per cent level. The report runs abstract, introduction ending in a stated hypothesis, method covering sample, ethics and data-collection tools, results with a scattergram, then discussion, and it is the discussion where marks sit, for third variables, self-report validity and improvements that actually follow from the weaknesses named.

The learning practical is observational, and the write-up has to carry both kinds of data from the same fieldwork. Decide first what counts as the behaviour, build a coding frame of categories that are observable and mutually exclusive, then tally by event or in fixed time intervals while a second observer tallies independently so that agreement can be reported. Field notes and remarks overheard give the qualitative strand, which is coded and grouped into themes supported by quotations. The tallies go into a contingency table for chi-squared, which asks whether two nominal variables, such as sex of pedestrian and whether the door was held open, are associated. Expected frequency for a cell is the row total times the column total divided by the grand total, and the calculated value must reach the critical value before the null hypothesis is rejected.

The method turns words into numbers and then asks what the numbers mean. Coding categories are fixed before reading and operationalised tightly enough that a second person would tally the same way, for instance every use of a violence word, a recovery word or a medical word. Sources are sampled on a stated basis with comparable amounts of text from each, and the tallies are set side by side. Because the data are frequencies in named categories drawn from independent sources, a chi squared test of association fits; degrees of freedom are rows minus one multiplied by columns minus one, so a table of two rows and two columns gives one, and the calculated value must reach or pass the critical value at the five per cent level before the null hypothesis is rejected. A second coder on part of the material gives inter rater reliability.

The write up is marked as a piece of research, so each section has a job. The aim narrows into a testable hypothesis with both variables operationalised, the method names the design and justifies it, sampling says how participants were obtained and what that does to generalisability, and ethics is settled in the plan rather than apologised for afterwards. The test follows three questions: difference or association, independent or related data, and the level of measurement. Frequencies in categories take chi squared, ordinal scores from two separate groups take Mann-Whitney U, the same people measured twice take Wilcoxon, and two scores from each person take Spearman's rho. The calculated value is then compared with the critical value and a decision stated about the null hypothesis.

Everything else follows from one decision, which is the test you will run, so make it before collecting anything. Frequencies of people in categories from independent participants go to chi squared, where an expected frequency is the row total times the column total divided by the grand total and the degrees of freedom are rows minus one times columns minus one. Ordinal scores from two separate groups go to Mann-Whitney, the same participants measured twice go to Wilcoxon, and two scores from each person that you want to relate go to Spearman’s rho. Write the hypothesis so it names difference or relationship and says whether it is directional, and settle consent, confidentiality and debriefing before anyone is approached.

The write-up is marked as a piece of research, so every decision has to be defensible and the analysis has to follow from the design. Attitudes to drinking gathered by questionnaire, accounts of quit attempts gathered by interview and drug references coded from media will all yield numbers provided the categories are fixed in advance. The test then follows from two questions: difference or relationship, and if difference, whether the same people appear in both conditions. Frequencies in categories take chi squared, ranked scores from two separate groups take Mann-Whitney, ranked scores from the same people twice take Wilcoxon, and paired scores take Spearman. Ethics carry real weight here because the subject is drug and alcohol use, so anonymity, consent and the right to withdraw are design features rather than paperwork.