Plan the research for your psychology master's thesis around a question, then choose the measures and analysis. First establish what you want to understand, whom you will study and which information can answer your question. Investigating an association between two characteristics calls for a different design from comparing two groups. Below, we will work through both approaches.
You have a topic involving stress, well-being or student life. You find several questionnaires, but still cannot see how they fit together as a study. Your supervisor asks about variables, participants and an analysis plan, and “I'll run a survey” is no longer a sufficient answer. Adding more scales will not resolve the problem. You need a coherent idea, from the psychological question to what the collected data will actually allow you to conclude.
The first example examines the association between perceived support and stress among students. The second compares stress among students who work and those who do not. These are different questions, even though both concern stress and a similar population. We will use them to explain decisions worth discussing before distributing your survey. This article does not report a completed study or diagnose particular individuals.
What should your psychology study actually help you understand?
“Student stress” identifies an area of interest, but not yet a research task. You might want to describe levels of stress, examine an association with support, compare groups or assess change over time. Each approach requires different data. Before choosing one, write a sentence beginning “I want to investigate whether…”. Then check whether you can identify the information needed to answer it.
In psychological research methods, the meanings of your concepts also matter. Support could mean believing that someone is available to help, the number of friends a person has or help actually received in the past month. These meanings are not interchangeable. In our first example, we choose perceived social support, rather than the size of a person's social network.
Jerzy Marian Brzeziński explains the relationship between theory, measurement, study design and interpretation. This is a useful warning against choosing a questionnaire simply because it is easy to obtain. The diagram appears on page 440 of his article in “Roczniki Psychologiczne”. First explain what a concept means in your thesis, then check whether the instrument actually measures it.
Example one: is social support associated with student stress?
Start with this question: “Is perceived social support associated with perceived stress among adult students at the selected university?” Each participant completes a measure of support and a measure of stress in the same study. This gives us two scores for each person. We want to know whether differences in one tend to accompany differences in the other.
From question to data: perceived support and perceived stress – two clearly defined variables – suitable versions of the instruments – two scores from the same participant – analysis of the association between them.
We are not yet asking whether support reduces stress. Measuring both characteristics once does not establish which came first. People going through a difficult period may perceive the availability of help differently, and other circumstances may also contribute to the association. For this design, a question about co-occurrence is more defensible than a title promising to demonstrate an effect.
How do you choose measures that fit the question?
For each instrument you are considering, note its full name, authors, Polish adaptation and intended scope. Check whether its score represents the characteristic described in your theoretical chapter. If a scale concerns stress during the past month, do not automatically describe it as measuring a stable capacity to withstand stress. The time frames of the two measures should also make sense within your project.
Read the documentation before building the survey. Establish how scores are calculated, which direction they run in and how missing responses should be handled. Decide whether you need an overall score or individual subscales. Adding every subscale “just in case” can create numerous additional questions that the topic does not justify.
This example does not require protected test items to be reproduced in the article. For more detail on describing an instrument, see How to describe a research instrument in your thesis: an original questionnaire and an existing scale. At this stage, the study plan needs a clear rationale: what information will each measure provide, and why is that information necessary to answer your question?
Correlation: what do you plan before seeing the responses?
The first approach starts with the association between two scores. Choose an appropriate coefficient with regard to the type of data, the shape of the relationship, outliers and the assumptions needed for inference. Pearson's correlation describes a linear relationship; Spearman's rank correlation concerns a monotonic relationship. An automatic rule such as “psychology always uses Spearman” cannot replace that assessment.
Before calculating the coefficient, inspect a scatterplot. You may see a curve, a distinct cluster of observations or a single entry that needs checking. A coefficient alone will not reveal all of this. Schober, Boer and Schwarte discuss the use and interpretation of correlation in a peer-reviewed article available through the Amsterdam UMC repository.
Make the plan conditional but specific: which data will you inspect, how will you assess whether the proposed analysis makes sense, and which information will accompany the result? Avoid saying “I will run all the correlations”. The main question concerns two particular scores, so start with that relationship. Additional exploratory work can be reported separately, with its purpose clearly identified.
Which conclusions will the design support?
If the analysis indicates an association, your conclusion should address its direction, size and uncertainty in the sample studied. “Higher support was associated with lower stress” means something different from “support protects against stress”. The latter requires a stronger basis than measuring two characteristics at the same time. You can compare your finding with theory in the discussion, but theory cannot change the design of your study.
If you do not obtain convincing evidence of an association, do not automatically conclude that support is irrelevant. Consider the precision of the estimate and the population to which the data relate. Rohrer's article on correlation and causal inference helps explain why simply adding more variables to an analysis does not resolve the problem of causality either.
Example two: do working and non-working students differ in stress?
The question now becomes: “Does perceived stress differ between students in paid work and students who are not working?” We want to compare stress scores in two groups. Changing a heading does not transform the first research question into this one. We must also establish how group membership will be determined and whether the comparison will be interpretable.
From question to data: clearly defined paid employment – two groups of different individuals – the same measurement of stress – a description of each group – a comparison of scores, including the size of the difference and its uncertainty.
A simple “working: yes/no” question may be too vague. One person works regularly, another completed a one-off assignment, and a third will start next week. Agree on the period covered by the question and how borderline situations will be handled. Record the criterion before recruitment. This prevents participants from being moved between groups only after their scores have been inspected.
What matters beyond the names of the groups?
Both groups should come from the population you have described and complete a comparable procedure. If working students are recruited during exams and non-working students during a quieter period, the difference becomes harder to interpret. The same applies if one group completes the questionnaire in a classroom and the other in different conditions that are not reported.
Consider year and mode of study or working hours if they are relevant to the question. Do not collect a lengthy background questionnaire without knowing how the information will be used. Differences in age, study workload or family circumstances may help explain the composition of the groups, but collecting this information does not remove every alternative explanation.
Do not split a continuous stress score into “low” and “high” merely to create additional groups. In this example, the groups follow from the question about paid work. Stress remains a measured score. For more on translating a concept into observable information, see Operationalising variables in a thesis: indicators, measurements and examples.
Comparing groups: why not start with the name of a test?
First establish whether you are comparing different individuals or the same people measured twice. In this project, working and non-working students form independent groups. Measuring the stress of the same students before and during exams would be a different design. These designs are not interchangeable, even if the survey and topic look similar.
For two independent groups, a test comparing means may be one option, but choosing it requires an assessment of the data and assumptions. A different type of outcome or a different question may require another method. Do not select a test solely from the number of participants or one value reported by software. The guide Statistical analysis in a master’s thesis: choosing tests and reporting results develops this practical issue.
A comparison should not end with the words “significant” or “not significant”. You also need descriptions of both groups, the size of the difference and its uncertainty, reported appropriately for the analysis used. Readers can then see what differs and by how much. A significance threshold cannot replace psychological interpretation.
A group difference is not yet an effect of paid work
In this project, students are not randomly assigned to start working or remain out of work. Participants are already in different life circumstances. A clear difference in stress therefore does not, by itself, establish that work caused it. You can describe the comparison and consider its possible explanations and limitations.
This matters for chapter titles too. “Stress among working and non-working students” fits this comparison better than “The effect of employment on mental health”. The second title adds both a causal claim and a broader outcome that a stress score alone does not cover. Matching your language to the data starts before you write the discussion.
An instrument found online is not necessarily ready to use
Before starting the study, establish the language version, conditions of use and qualifications required to administer the instrument. Check permission for online administration, modifications and publication of materials separately. Finding a PDF is not, in itself, a licence. Obtain information about the particular scale from its author or publisher, with your supervisor's support.
Do not shorten an instrument or translate its items simply to make the survey more convenient. Changing the measurement procedure may change the meaning of the score. The International Test Commission statement in an authorised Polish translation discusses permission, modifications, test security and researcher responsibility. Apply the conditions governing the instrument you choose, rather than extending one publisher's rules to every test.
Both examples concern research findings, not the diagnosis of individual students. Do not promise participants a diagnosis or individual interpretation that the project does not provide. Participant information should reflect what will actually happen. Before recruitment, discuss your university's requirements for ethical review and documentation with your supervisor.
Whom should you recruit, and how do you justify the sample size?
Describe eligibility so that another person can understand who is being invited. “Students” is insufficient if the link will reach only friends from one year group. Establish inclusion criteria, recruitment channels and the planned data-collection period. In the group-comparison example, you also need a realistic way to recruit both groups, rather than simply enough responses overall.
The invitation may also affect who responds. An advert addressed to “very stressed students” encourages a different response from a neutral announcement about student functioning. Keep a record of the invitation and where it was shared. Do not call the sample representative merely because the link appeared in several online groups. A recruitment description helps readers judge to whom the findings may cautiously apply.
No single sample size guarantees a sound master's thesis. The rationale depends on the question, planned analysis, desired precision and available resources. Lakens distinguishes several approaches to justification instead of proposing one threshold for every study. His open-access article “Sample Size Justification” is useful reading before recruitment begins.
If resources are limited, state the limitation and examine which conclusions the planned data can support. Do not invent a retrospective explanation that the number recruited “is always sufficient”. If you need support, see the service “Statistical analysis for bachelor’s and master’s theses”. For the discussion, prepare your research questions, measurement plan and information about the population you can realistically reach.
Work through the study as a participant would
Before sharing the invitation, check the entire permitted procedure: participant information, consent, instructions, transitions between sections and the ending. Complete a trial survey in conditions matching its intended use. Check whether missing responses can be identified and whether the instrument retains the format required by its documentation. Convenience should not change the measurement procedure.
In the first example, check particularly that both scores belong to the same person and that an incomplete stress section can be recognised. In the second, make sure the employment response allows you to apply the agreed group criterion. A situation such as “I worked previously, but not during the period covered by the question” must be identifiable before analysis.
A procedural trial does not establish an instrument's validity or reliability. If participants find an item in a protected scale unclear, do not rewrite it yourself. Discuss the problem with your supervisor and the person responsible for the instrument. If you revise your own employment question, retain the version and date of the change. Also decide whether trial responses will be kept separate from the main dataset.
How do you describe a plan without promising a result?
A study plan should not decide in advance what the analysis will find. For the first approach, you might write: “Perceived social support and perceived stress will be measured in the same participants. The analysis will examine the association between the scores. Data characteristics and the assumptions of the proposed method will be assessed before selecting the inferential approach.” This is a starting point for refining the document, not a ready-made description for every project.
For the second approach, the essential information differs: “The comparison will involve two groups of independent participants defined by a predetermined criterion for paid employment. Stress will be measured in the same way in both groups. The result will be interpreted as a difference between the groups studied, taking account of recruitment limitations.” You must then specify the actual criterion, population and analysis in your own plan.
Notice which expressions are absent: “the study will prove”, “the result will confirm” and “working students will turn out to be more stressed”. An expectation derived from the literature may justify a hypothesis, but it is not a finding. In your notes, distinguish the prediction from the procedure, and the procedure from what the data may eventually show. An unexpected result then requires no rewriting of the study's history.
What should you record before distributing the study link?
Prepare a short, dated version of the plan. You do not have to begin with dozens of pages of methodology, but the document should let you check whether the decisions fit together. Go through it with your supervisor and resolve contradictions before participants start responding.
- Main question and scope. State whether you are describing an association, a difference or change. Record what the project does not test as well.
- Participants and recruitment. Establish inclusion criteria, access to the groups and the actual conditions of data collection.
- Measurement and procedure. Identify instrument versions, administration consistent with their documentation and the sequence of activities.
- Analysis. Specify the main outcomes, the rationale for the method and how uncertainty will be reported. Separate primary from additional analyses.
- Data-handling rules. Establish how missing responses, ineligible entries and possible duplicates will be addressed, as far as they can be identified.
These decisions reduce the temptation to select only the most interesting-looking results after the event. You can document them with your supervisor and, where required or planned, in a preregistration. Later changes may be necessary. Retain a record of what changed, when and why, rather than presenting a new decision as the original intention.
How do you connect methods, results and discussion?
Take the main question through the three parts of the thesis. The methods should explain how the necessary data were obtained. The results should answer the question using an appropriate analysis. The discussion should explain what the answer means in light of the literature and limitations. If a different variable appears in the results, clarify the discrepancy before merely polishing the wording.
In the first example, readers should be able to locate definitions of support and stress, descriptions of both measures and the finding concerning their association. In the second, they also need clear group criteria and an account of the comparison. A good connection between chapters does not mean repeating paragraphs. It means retaining the same question and the same meanings of the concepts.
If you already have data, do not rewrite the study's history as though it had followed a different design from the outset. Report the procedure that actually took place and discuss deviations. Through the service “Analysis and presentation of research results for your thesis”, I can help organise summaries and reporting based on data actually collected. I do not add measurements that never took place. Keep the distinction between a research hypothesis and the outcome of testing it; see Research hypothesis vs statistical hypothesis: planning a thesis analysis
What should you bring to a consultation to move forward?
Before data collection, prepare your topic, a short rationale, research questions and documentation for the instruments under consideration. Add whom you can reach and what you have agreed with your supervisor. If responses have already been collected, include an anonymised dataset, explanations of the columns and an account of what actually happened. Do not send participant names or test materials you are not authorised to share.
The Psychology page describes support for work in this discipline. The most useful message identifies your stage and difficulty: “I have a question about the association between support and stress, and two possible instruments” or “I have collected data from two groups, but my methods chapter does not describe what was done.” You do not need to choose a statistical test on your own beforehand.
Psychology thesis support: who is Student Support a good fit for?
Student Support is a good fit when you need to connect a psychological question, measurement and data analysis, rather than receive calculations alone. I have run the business since 2018. You discuss your topic, university requirements and available material directly with me, Wioletta Boćkowska. We can start with your research plan or with data you have already collected, and match the support to the actual stage of your thesis.
If your project requires your own survey questions, consider Questionnaire design for your thesis. We distinguish this work from using standardised tests, which require the appropriate version, conditions of use and interpretation. If you need help with connected parts of the empirical work, choose Thesis research section: methodology, results and discussion. Its agreed scope can bring together the methodology, your actual material and the presentation of results.
If you first need to discuss a decision rather than commission work on a chapter, Bachelor’s and master’s thesis consultations is an appropriate option. Well-matched support should help you understand how the analysis relates to your research question and explain the limits of your conclusions. At Student Support, we define the scope around your project and confirm the price and deadline before work begins.
You can also commission a model bachelor’s or master’s thesis in psychology, or a specified chapter, through Bachelor’s and master’s thesis writing. I tailor the scope to your topic, outline and university requirements. If the work includes an empirical section, we establish which instruments, data and materials are actually available. You receive a proposed scope, price and deadline before work begins.
References and further reading
- Brzeziński, J. M. (2016). O osobliwościach metodologicznych badań naukowych i diagnostycznych prowadzonych przez psychologów klinicznych. Roczniki Psychologiczne, 19(3), 437-452. https://doi.org/10.18290/rpsych.2016.19.3-2pl
- International Test Commission. (2014). Stanowisko ITC w sprawie stosowania testów i innych narzędzi oceny w badaniach naukowych (version 1.2; Polish translation authorised by the Polish Psychological Association).
- Lakens, D. (2022). Sample size justification. Collabra: Psychology, 8(1), Article 33267. https://doi.org/10.1525/collabra.33267
- Rohrer, J. M. (2018). Thinking clearly about correlations and causation: Graphical causal models for observational data. Advances in Methods and Practices in Psychological Science, 1(1), 27-42. https://doi.org/10.1177/2515245917745629
- Schober, P., Boer, C., Schwarte, L. A. (2018). Correlation coefficients: Appropriate use and interpretation. Anesthesia & Analgesia, 126(5), 1763-1768. https://doi.org/10.1213/ANE.0000000000002864
These sources help establish the foundations of planning, measurement and interpretation. Consult the documentation for your chosen instruments and current university thesis requirements separately. The article explains the reasoning through two examples; it does not replace a study plan agreed with your supervisor.
Questions about psychology research and reporting results
Can I revise the methodology of a psychology thesis after data collection?
Yes. Please send the topic, research questions, hypotheses, a description of the instruments, and information about the actual course of the study. I will help you organise the methodology according to what was actually done and point out any discrepancies between the assumptions and the description.
What should I prepare for the analysis and presentation of results in a psychology thesis?
Include the research questions, hypotheses, description of variables and tools, and available analysis summaries. If you also need calculations, send an anonymised dataset with a description of columns and the coding method for responses. Based on this, I will determine the scope of support.
Do I need to know which statistical test to commission?
No. Not every thesis requires statistical tests. Please send the research questions, hypotheses, the survey, and the database. I will check the type of variables, group structure, and method assumptions, and then propose an analysis that corresponds to the research objective.
How many people are needed for statistical analysis in a master's thesis?
There is no single number suitable for every thesis. The planned test, the number and size of groups, the expected effect size, and the required power and precision of the analysis are important. A small sample limits inference. It is best to discuss the sample size before starting data collection.
Let's plan the next step in your psychology thesis
Send your topic, questions, supervisor's comments and a description of your research stage. I can help identify whether you need to refine the methods, analyse the data or report the results. If data have already been collected, include an anonymised copy with explanations of the columns. We will discuss scope, price and deadline before work begins.
