You have the analysis report for your psychology thesis. The test was run, but the result was statistically nonsignificant at the chosen threshold. Your supervisor asks for two separate paragraphs, one for Results and one for Discussion. What should you write so the finding is visible without claiming that the effect does not exist?
In Results, say what you compared, how many observations you analysed, how large the estimated effect was, which direction it pointed in and how precise it was. In Discussion, explain whether the analysis supported your prediction, what remains plausible and what limits your conclusion. Knowing that p exceeded the threshold is too little for either paragraph.
This guide focuses on the final stage of analysis in Psychology. If you are still preparing the dataset or checking how to score your scales, start with Statistical analysis for a psychology master’s thesis – what to do with collected data. Here the calculations are already done and the job is to report them accurately.
Look for the estimated effect in your report
Return to the research question and the hypothesis as they stood before you saw the results. In one sentence, name the comparison and the order of subtraction. For repeated measurements on the same people, that might be “after minus before”. For independent groups, it might be “group A minus group B”. In a correlation, report the estimate of the relationship between the two variables. Reversing the sign changes the meaning of the sentence even when the two-sided p value stays the same.
- Observations analysed. Record the number used in this test. In a before-and-after comparison, count the complete pairs. The total number of measurements answers a different question.
- Effect size. Choose the relevant estimate and give its units. A difference of 1.8 scale points is in the original units. A standardised effect size answers a different question and should be defined if your report requires one.
- Uncertainty and test. Note the confidence level and interval limits, the name of the test, its statistic, degrees of freedom where applicable and the p value. State the threshold set in advance and check whether the test was one-sided or two-sided.
If the test could not be run, for instance because no participant has a second measurement, say what prevented you from assessing the hypothesis. That is different from running the test and obtaining a result above the threshold. APA JARS-Quant provides a useful reporting checklist. Follow the format and notation agreed with your supervisor and required by your university in Poland.
Example: stress before and after a study preparation routine
This example is fictional. Twenty-four people completed the same self-report stress scale at the start and after two weeks of using a short routine to prepare for studying. Scores were rescaled to a range of 0 to 40 and could include fractions. Higher scores indicate more self-reported stress. The prediction was that the average score would be lower at the second measurement. The plan specified a two-sided paired t test and α = 0.05. There was no comparison group.
Follow the numbers. For each person, the baseline score was subtracted from the two-week score. The mean difference was -1.80 points and the standard deviation of the differences was 5.40. The standard error was about 1.10. With 24 pairs independent across participants, this gives t(23) = -1.63, two-sided p = 0.116 and a 95% confidence interval for the mean change from -4.08 to 0.48 points. The example assumes independence across participants and an approximately normal distribution of within-person differences. Use your own study design and data to assess these conditions.
A model Results paragraph
Change in self-reported stress between the first and second measurements was assessed with a two-sided paired t test on 24 complete pairs. The mean difference (second minus first measurement) was -1.80 points, 95% CI [-4.08, 0.48], t(23) = -1.63, p = 0.116. At α = 0.05, the result was statistically nonsignificant.
The direction and units appear before the p value. If you already have a summary of sample sizes and descriptive statistics for both measurements, refer to it in the chapter as your university requires. The example illustrates one comparison. In your thesis, report the other planned tests and any adjustments for multiple comparisons that apply to your analysis plan.
Keep three conclusions separate in the Discussion
- Direction in the sample. The estimated mean change was -1.80 points. In this sample, the second score was lower on average.
- Precision of the inference. The interval includes zero, average decreases of up to about 4.08 points and a small increase of up to about 0.48 points. Under the method and its assumptions, the estimate is too imprecise to separate these values clearly. Describe them as values compatible with the data under the model. The 95% level refers to the long-run behaviour of the procedure used to construct the intervals.
- What the design can establish. Without a comparison group, any change cannot be attributed to the routine. Course workload, measurement timing or other experiences could have changed over the two weeks. The example did not test these explanations, so they remain possibilities for future research.
A model Discussion paragraph
The analysis provided insufficient support for the prediction of lower mean stress at the second measurement. The estimate pointed to a decrease in the sample, but the uncertainty was consistent with either no average change or a decrease of several points. With two measurements and no comparison group, the study cannot establish whether any change was associated with the routine or with other changes during that period.
When comparing your finding with the literature, compare what was actually measured. A controlled study may estimate the difference in change between groups, while a design without a control estimates change in one group. Compare the populations, the time between measurements, the instruments and the size and precision of the effects. Different labels of statistical significance do not, by themselves, establish different effects. For the wider chapter structure, see How to write the discussion in a bachelor’s or master’s thesis.
When can you call a difference practically small?
What a scale point means depends on the instrument and your research question. A claim that the effect falls within a practically negligible range needs justified bounds and a method suited to that question, such as the two one-sided tests (TOST) approach to equivalence. Ideally, set the bounds before looking at the result, using theory, earlier data or the meaning of the scale. Exceeding the threshold in a standard test for a difference does not support that conclusion. The fictional study set no equivalence bounds and ran no equivalence test.
The ASA notes that the p value does not measure either effect size or the probability that a hypothesis is true. The sentence “p = 0.116 means there is an 11.6% chance that there is no effect” assigns a meaning the statistic does not have. To discuss practical relevance, use the estimate, units, precision and quality of the research design.
What to do if you find an error after analysis
It is reasonable to check a possible mistake. If you find mismatched participant IDs or a scale item that was reverse-scored incorrectly, keep the earlier report, correct the error, rerun the analysis and document the change. Explain what changed and why. A justified sensitivity analysis can also help if you state its purpose and report it alongside the primary analysis.
Keep the threshold, direction of testing, exclusion rules and outcome definition independent of any attempt to get a smaller p value. You can investigate a question that came up after seeing the data and label it exploratory. Keep the original test in the chapter. Readers can then distinguish the planned analysis from ideas that came later.
Check the paragraph before submission
- Do the test name, number of observations and sign of the estimate match the report?
- Does the interval refer to the same quantity as the estimate, with the units stated?
- Does the Discussion explain the limit of the conclusion instead of repeating all the statistics?
- Have you kept the primary analysis and clearly labelled additional analyses?
Questions about hypotheses, confidence intervals and discussion
What if the results do not support the hypothesis?
I describe this in line with the analysis. A lack of support for the hypothesis is a research finding, not a reason to change the respondents' answers.
It also does not automatically mean the aim was not achieved: the aim may have been to check a specific relationship. I explain what follows from the data and what limitations must be considered. A lack of support for the hypothesis does not by itself prove that the relationship definitely does not exist.
How should I describe a confidence interval in a master's thesis?
State the confidence level, the estimated parameter, and the interval limits, for example: the mean difference was 4 points, and its 95% confidence interval ranged from 1.33 to 6.67 points. Maintain the unit of measurement and specify for which comparison the interval was calculated.
In a frequentist interpretation, the 95% level refers to the procedure: in repeatedly conducted studies meeting the assumptions, approximately 95% of such intervals would contain the true parameter. It does not mean that 95% of participants obtained a result between the given limits. A wide interval indicates lower precision of the estimate.
What materials should I send for editing the discussion section of a psychology thesis?
Send the aim of the work, questions and hypotheses, the results chapter, the methodology description, and the literature used. Include the existing discussion and the supervisor's comments, if you have them. These materials allow the interpretation to be related to the actual results and previous studies.
How does preparing the research chapter differ from statistical analysis?
The research chapter combines a description of the actual study, presentation of results, interpretation and answers to the thesis research questions. Statistical analysis concerns the choice of methods and calculations on quantitative data. You can request both types of support together, but I agree any new tests and verification of calculations separately after reviewing the material.
References
- Wasserstein, R. L., & Lazar, N. A. (2016). The ASA's Statement on p-Values: Context, Process, and Purpose. The American Statistician, 70(2), 129-133. DOI: https://doi.org/10.1080/00031305.2016.1154108
- Lakens, D. (2017). Equivalence Tests: A Practical Primer for t Tests, Correlations, and Meta-Analyses. Social Psychological and Personality Science, 8(4), 355-362. DOI: https://doi.org/10.1177/1948550617697177
- Appelbaum, M., Cooper, H., Kline, R. B., Mayo-Wilson, E., Nezu, A. M., & Rao, S. M. (2018). Journal Article Reporting Standards for Quantitative Research in Psychology: The APA Publications and Communications Board Task Force Report. American Psychologist, 73(1), 3-25. DOI: https://doi.org/10.1037/amp0000191
Let’s work out what support you need
Tell us what you already have: data, an analysis report or draft paragraphs. Aggregate results are enough for the first message. Include your university’s guidance and say whether you need the analysis checked, help reporting the results or work on the discussion.
