SPSS for Dissertations: A Step-by-Step Guide for Beginners | Writers.ng

How to Write a Dissertation Introduction in UK Universities

For many students, the first encounter with SPSS is a grey screen full of rows and columns, a menu that seems to offer a thousand options, and a supervisor who says “just run the right test”. It is no wonder that statistics causes more dissertation anxiety than almost anything else.

The good news is that SPSS is a tool with a manageable core. You do not need to master every menu. You need to know how to set up your data properly, check and clean it, choose a test that fits your question and data, run it, read the output and report the results correctly. This guide walks through each step with the menu paths, examples and reporting templates most UK dissertations require. For support across the whole project, see our main guide to dissertation help UK.

What is SPSS and why do UK universities use it?

SPSS (Statistical Package for the Social Sciences, now IBM SPSS Statistics) is statistical software with a point-and-click interface. It is widely available through UK university software catalogues, common in psychology, business, education, health and social science, and well supported by teaching resources. The publisher’s IBM SPSS Statistics pages describe current versions and features.

Table of Contents

Alternatives include R, Stata, Jamovi and JASP. If your department prefers one of these, follow its guidance. The statistical thinking in this guide applies whichever software you use, though the menu paths below are for SPSS and may differ slightly between versions.

Before you open SPSS: plan your analysis

The biggest mistake students make is collecting data first and wondering about analysis later. Before you collect anything, decide:

  1. Your research question and hypotheses. What exactly are you testing?
  2. Your variables and their types (see below).
  3. The test you are likely to use for each hypothesis.
  4. The sample size you need to run that test with adequate power.

Your methodology chapter should link every research question to a planned analysis. If you are still shaping your design, read our guides to the dissertation methodology chapter and qualitative vs quantitative dissertations.

Understand your variables

Choosing the right test starts with knowing what kind of data you have.

Level of measurementMeaningExamplesSPSS “Measure” setting
NominalCategories with no orderGender, department, yes/noNominal
OrdinalOrdered categories, unequal gapsSatisfaction rating (low to high), education levelOrdinal
Interval or ratioNumeric with equal gapsAge, test score, income, reaction timeScale

Likert scales are a common source of confusion. A single Likert item is technically ordinal. A composite score built from several items (for example the mean of ten items) is often treated as scale data, but conventions vary by field, so follow your department’s guidance and justify your choice.

Also identify your independent variable (the predictor, group or condition) and dependent variable (the outcome).

Step 1: Set up your data file

SPSS has two main views, accessed from tabs at the bottom left.

  • Data View: a spreadsheet where each row is a participant (or case) and each column is a variable.
  • Variable View: where you define each variable’s properties.

Set up Variable View first

For each variable, fill in:

ColumnWhat to enterTip
NameShort, no spaces (for example Age, Q1, JobSat)Use consistent names; SPSS does not allow spaces
TypeNumeric for most data; String for textAvoid string variables for analysis
LabelA full description (“Job satisfaction, item 1”)Appears in output, so write clearly
ValuesCodes for categories (1 = Female, 2 = Male, 3 = Other)Use the Value Labels dialog
MissingCodes for missing data (for example 99)Tell SPSS so it ignores them
MeasureNominal, Ordinal or ScaleSet correctly; some procedures depend on it

Enter or import data

  • Type directly into Data View for small datasets.
  • Import from Excel or CSV: File, then Open, then Data, choose the file type, and tick “Read variable names from the first row of data”.
  • Import from survey platforms. Many tools export directly to SPSS (.sav) format or to Excel with codes.

If you are using an existing dataset, such as those from the UK Data Service, read the documentation carefully for variable definitions, weighting and licensing conditions.

Keep a data dictionary and a backup

Record what each variable means and how it is coded. Save your data file frequently with clear version names, and keep a separate backup. Also save your syntax (the code SPSS generates) via the Paste button in dialogs. This creates a record of every step and lets you repeat your analysis exactly.

Step 2: Clean and check your data

Bad data produce unreliable results, and markers notice. Allow real time for this step.

Check for entry errors

Run Analyze, then Descriptive Statistics, then Frequencies for categorical variables and look for impossible values (a gender code of 7, an age of 250). Run Descriptives for numeric variables and check minimums and maximums.

Handle missing data

  • Check how much is missing and whether it is random.
  • Decide how to deal with it and justify the choice: leave it out (listwise or pairwise deletion), or in some cases use more advanced methods under supervision.
  • Report how many cases and variables were affected.

Reverse-code negatively worded items

If your scale includes items such as “I often feel unproductive” alongside positive items, reverse them so that high scores mean the same thing throughout. Use Transform, then Recode into Different Variables, create new variables rather than overwriting the originals, and label them clearly.

Create composite scores

Use Transform, then Compute Variable to calculate a total or mean, for example the mean of items Q1 to Q10. Check how missing items are handled (the MEAN function with a minimum number of valid items is often better than a simple sum).

Check for outliers

Use boxplots (Graphs, then Chart Builder, or Analyze, then Descriptive Statistics, then Explore) to spot extreme values. Investigate before removing anything. Outliers may be errors, or genuine and informative. Report what you did.

Split or select cases (if needed)

Use Data, then Select Cases to analyse a subgroup, and remember to turn the filter off afterwards. Forgetting this is a classic source of wrong results.

Step 3: Describe your sample

Every quantitative dissertation needs a participant or sample description.

  • Categorical variables (gender, department): Analyze, then Descriptive Statistics, then Frequencies. Report counts and percentages.
  • Numeric variables (age, tenure): Analyze, then Descriptive Statistics, then Descriptives (or Explore). Report mean, standard deviation, minimum and maximum.

Present this in a table, with clear labels. If you have grouped data, compare descriptives across groups to give the reader a feel for the pattern before tests.

Step 4: Check the reliability of your scales

If you use multi-item scales, you should report their internal consistency, usually with Cronbach’s alpha.

Menu path: Analyze, then Scale, then Reliability Analysis. Move the items into the box, choose Alpha as the model, and under Statistics tick “Scale if item deleted”.

Reading the output:

  • Alpha values of about .70 or above are commonly treated as acceptable for research purposes, though conventions vary and higher is expected for some uses.
  • Check Corrected Item-Total Correlation for weak items and Cronbach’s Alpha if Item Deleted to see whether removing an item would improve reliability. Only drop items with good reason, and report it.

Reporting example (illustrative): “The job satisfaction scale showed good internal consistency (Cronbach’s α = .86).”

If you adapted or translated an existing scale, report the reliability in your sample even if the original scale reported it too.

Step 5: Choose the right test

This is the step that causes the most worry. Use your question and variable types to narrow down the options.

Your questionVariablesCommon parametric testCommon non-parametric alternativeSPSS path
Do two independent groups differ on a scale outcome?1 categorical (2 groups), 1 scaleIndependent-samples t-testMann-Whitney UAnalyze, Compare Means, Independent-Samples T Test; or Nonparametric Tests
Do scores change between two time points in the same people?1 scale measured twicePaired-samples t-testWilcoxon signed-rankAnalyze, Compare Means, Paired-Samples T Test
Do three or more groups differ on a scale outcome?1 categorical (3+ groups), 1 scaleOne-way ANOVAKruskal-WallisAnalyze, Compare Means, One-Way ANOVA
Is there a relationship between two scale variables?2 scalePearson correlationSpearman correlationAnalyze, Correlate, Bivariate
Is there an association between two categorical variables?2 categoricalChi-square test of independenceFisher’s exact (small counts)Analyze, Descriptive Statistics, Crosstabs
Can one or more predictors explain a scale outcome?1 scale outcome, 1+ predictorsLinear regressionRobust or other methodsAnalyze, Regression, Linear
Can predictors explain a binary outcome?1 binary outcome, 1+ predictorsLogistic regressionn/aAnalyze, Regression, Binary Logistic
Is a scale reliable?Multiple itemsCronbach’s alphan/aAnalyze, Scale, Reliability Analysis

If your design is more complex, such as repeated measures with several factors, mixed designs, factor analysis or multilevel models, speak to your supervisor or a methods adviser. Our dissertation data analysis help page explains how we support students in understanding and running their own analysis.

Step 6: Check assumptions

Parametric tests rest on assumptions. Checking them shows rigour and prevents misleading results.

AssumptionApplies toHow to check in SPSSIf violated
Normalityt-tests, ANOVA, correlation, regression residualsExplore with Normality plots and tests; histograms and Q-Q plots; skewness and kurtosisUse a non-parametric alternative or transform, with justification
Homogeneity of varianceIndependent t-test, ANOVALevene’s test (in output)Use the corrected row (equal variances not assumed) or Welch ANOVA
Independence of observationsMost testsDesign checkRedesign or use appropriate methods
LinearityCorrelation, regressionScatterplotsConsider non-linear terms or other methods
No multicollinearityMultiple regressionVIF and Tolerance under StatisticsRemove or combine predictors
HomoscedasticityRegressionPlot of standardised residuals against predicted valuesConsider transformation or robust methods
Expected countsChi-squareFootnote under the table: cells with expected count below 5Combine categories or use Fisher’s exact test

With larger samples, formal normality tests (such as Shapiro-Wilk) often flag trivial departures, so use them alongside plots and judgement, and follow your department’s conventions. Report what you checked and what you decided.

Step 7: Run and interpret the main tests

Below are the most common tests in student dissertations, with menu paths, what to look for in the output and how to report. All numbers in the reporting examples are invented for illustration. Never copy them into your own work.

Independent-samples t-test

Use when: comparing the means of two separate groups on a scale outcome.

Menu: Analyze, Compare Means, Independent-Samples T Test. Put the outcome in Test Variable(s), the grouping variable in Grouping Variable, then Define Groups.

Read: Levene’s test first. If Sig. is above .05, use the “Equal variances assumed” row; otherwise use “Equal variances not assumed”. Then read t, df and Sig. (2-tailed). Add the effect size (Cohen’s d, available in newer versions, or calculated by hand).

Report (illustrative): “Employees working hybrid (M = 3.9, SD = 0.7) reported higher engagement than employees working fully on site (M = 3.4, SD = 0.8), t(118) = 3.52, p < .001, d = 0.64.”

Paired-samples t-test

Use when: the same people are measured twice.

Menu: Analyze, Compare Means, Paired-Samples T Test.

Report (illustrative): “Anxiety scores decreased from pre-intervention (M = 24.1, SD = 5.2) to post-intervention (M = 20.3, SD = 4.9), t(39) = 4.10, p < .001.”

One-way ANOVA

Use when: comparing the means of three or more groups.

Menu: Analyze, Compare Means, One-Way ANOVA. Under Options tick Descriptive, Homogeneity of variance test and, where relevant, Welch. Under Post Hoc choose a test such as Tukey (equal variances) or Games-Howell (unequal variances).

Read: the ANOVA table (F, df, Sig.). If significant, check the post hoc table to see which groups differ. Report an effect size such as eta squared.

Report (illustrative): “There was a significant difference in motivation across the three working arrangements, F(2, 147) = 5.31, p = .006, η² = .07. Post hoc comparisons using Tukey’s test indicated that…”

Pearson correlation

Use when: examining the strength and direction of a linear relationship between two scale variables.

Menu: Analyze, Correlate, Bivariate. Tick Pearson, Two-tailed and Flag significant correlations. Always make a scatterplot as well.

Read: the correlation coefficient (r) and Sig. Interpret size cautiously and in context (conventions of about .10 small, .30 medium and .50 large are widely cited, but context matters). Remember that correlation is not causation.

Report (illustrative): “Job satisfaction was positively correlated with engagement, r(148) = .52, p < .001.”

Chi-square test of independence

Use when: testing association between two categorical variables.

Menu: Analyze, Descriptive Statistics, Crosstabs. Put one variable in Rows and one in Columns. Under Statistics tick Chi-square and Phi and Cramer’s V. Under Cells tick Observed and Expected counts, and row or column percentages.

Read: check the footnote for expected counts below 5, then the Pearson Chi-Square line.

Report (illustrative): “There was a significant association between working arrangement and intention to stay, χ²(2, N = 150) = 8.41, p = .015, Cramer’s V = .24.”

Simple and multiple linear regression

Use when: predicting a scale outcome from one or more predictors.

Menu: Analyze, Regression, Linear. Put the outcome in Dependent and predictors in Independent(s). Under Statistics tick Estimates, Confidence intervals, Model fit, R squared change, Collinearity diagnostics. Under Plots request standardised residual plots.

Read:

  1. Model Summary: R squared (the proportion of variance explained).
  2. ANOVA table: whether the model as a whole is significant.
  3. Coefficients table: unstandardised B, standardised Beta, t and Sig. for each predictor, plus collinearity statistics.

Report (illustrative): “The model significantly predicted engagement, F(3, 146) = 18.2, p < .001, R² = .27. Autonomy (β = .31, p < .001) and recognition (β = .24, p = .004) were significant predictors, while workload was not (β = −.09, p = .21).”

For binary outcomes, use binary logistic regression and report odds ratios with confidence intervals. For categorical predictors, create dummy variables or use the Categorical option, and be clear about reference categories.

Step 8: Understand p-values, effect sizes and confidence intervals

Markers increasingly expect more than “p < .05”.

  • p-value: the probability of results at least as extreme as yours if the null hypothesis were true. It does not tell you the probability that your hypothesis is true or how big or important an effect is.
  • Effect size: how large the difference or relationship is (Cohen’s d, eta squared, r, Cramer’s V, R squared, odds ratios). Always report one where possible.
  • Confidence interval: a range of plausible values for the effect. Report it where SPSS provides it.
  • Statistical vs practical significance: with a large sample, tiny effects can be significant. With a small sample, meaningful effects can be non-significant. Discuss both.
  • Non-significant results are results. Report them honestly and discuss possible reasons, including power.

Avoid language such as “proved” or “trend towards significance”. Use cautious wording such as “the results suggest” or “there was no statistically significant difference”.

Step 9: Report your results clearly

A good results section is clear, organised and honest.

Structure

  1. Data screening and preparation: missing data, outliers, assumption checks.
  2. Sample description: a table of descriptives.
  3. Reliability of measures: Cronbach’s alpha for each scale.
  4. Results by research question or hypothesis, in the same order as in your introduction and methodology.
  5. Summary: brief, linked to the discussion.

Tables and figures

  • Give each a number and a descriptive title, and refer to it in the text.
  • Follow your department’s referencing style, often APA 7 for statistical reporting. Cite Them Right is a trusted guide to citation formats.
  • Do not paste raw SPSS output. Rebuild clean tables with the key statistics.
  • Use consistent decimal places (usually two, with p-values to two or three decimals, and “p < .001” for very small values).
  • Keep the text for interpretation; do not repeat every number from the table.

Appendix

Place full output, syntax and your data dictionary in an appendix if your department allows, which supports transparency. Check anonymity first.

Common mistakes in SPSS dissertations

  • Choosing the test after seeing the data to get a significant result.
  • Wrong measurement level, such as treating nominal variables as scale.
  • Not setting missing values, so “99” is treated as a real score.
  • Forgetting to reverse-code negatively worded items.
  • Ignoring assumptions, or checking them but not reporting what you did.
  • Running many tests without considering multiple comparisons.
  • Reporting only p-values, with no effect sizes or descriptives.
  • Interpreting correlation as causation.
  • Leaving Select Cases on and analysing only a subset by accident.
  • Overwriting original variables instead of creating new ones.
  • Copying SPSS output tables straight into the dissertation.
  • Too small a sample for the planned analysis, such as regression with many predictors.
  • Not saving syntax or backups, then being unable to reproduce results.
  • Confusing statistical and practical significance.

Ethics, data protection and integrity

Quantitative data often include personal or sensitive information. Make sure you:

  • Follow the consent and confidentiality commitments in your ethics approval.
  • Anonymise data files, removing names, emails and direct identifiers.
  • Store files securely, in line with UK GDPR and the Data Protection Act 2018.
  • Follow your retention and deletion plan.
  • Report all analyses honestly, including those that did not support your hypotheses.

Useful references include the British Psychological Society, the BERA ethical guidelines, the UK Research Integrity Office and, for health research, the NHS Health Research Authority. See also our guide to dissertation ethics approval in the UK.

Learning SPSS efficiently

  • Use your university’s resources. Many run SPSS workshops, drop-in clinics and online tutorials.
  • Practise on a sample dataset before using your own.
  • Work in small steps, and check the output after each.
  • Use syntax from the start so every step is recorded.
  • Keep a log of decisions: why you chose a test, how you handled missing data, what changed.
  • Ask early. A methods adviser can save weeks if you ask before collecting data.
  • Learn the logic, not just the clicks. Understanding why a test fits matters more than memorising menus.

For guidance on productive conversations with your supervisor about analysis, see working with your dissertation supervisor.

SPSS advice for international and Nigerian students

Many students from Nigeria and across Africa study with UK universities, in person or through transnational programmes. A few points are worth noting.

  • Check software access early. Remote access to university SPSS licences can be limited by connectivity, so plan to install or connect in advance and keep local backups.
  • Document your data source carefully if you use Nigerian or African secondary datasets, including coverage, definitions and licensing.
  • Plan for response rates when distributing surveys across borders, and explain any sampling limits.
  • Check scale validity in your context. If you adapt a scale developed elsewhere, report how you checked that items were understood as intended and report reliability in your sample.
  • Be explicit about analytic reasoning. UK markers reward justification, not just output.

Our online dissertation support services are designed with these situations in mind. If your project is qualitative or mixed, see our thematic analysis and NVivo guide.

Academic integrity and your analysis

Your analysis must be your own, based on your own data, and reported honestly. Never alter data to obtain a result, and do not present output you do not understand. Learning from tutorials, model examples and feedback is sensible, as long as you follow your university’s rules on outside help. For more, see how to avoid plagiarism in your dissertation and the QAA’s UK Quality Code for the standards UK institutions work to.

SPSS dissertation checklist

  1. I planned my analysis before collecting data.
  2. Every variable is defined in Variable View with correct type, labels, values and measure.
  3. Missing values are declared and handled transparently.
  4. I checked for entry errors and outliers.
  5. Negatively worded items are reverse-coded.
  6. I reported reliability for multi-item scales.
  7. I chose each test based on my question and variable types.
  8. I checked and reported assumptions.
  9. I report test statistics, degrees of freedom, p-values and effect sizes.
  10. I interpret results cautiously, without implying causation.
  11. Tables and figures are clean, numbered and referenced in the text.
  12. I saved syntax, backups and a data dictionary.
  13. Data are anonymised and stored securely.
  14. Non-significant results are reported honestly.

Frequently asked questions

Is SPSS hard for beginners?

It is easier than it looks. The interface is menu-driven, and most dissertations rely on a small set of procedures. The harder part is choosing appropriate tests and interpreting results, which is a matter of understanding rather than software skill.

Which statistical test should I use for my dissertation?

It depends on your research question, the number and type of variables and your design. Use the table in Step 5 as a starting point, and confirm your plan with your supervisor.

What sample size do I need?

It depends on the test, the expected effect size and the power you want. Use a power analysis (for example with free software such as G*Power) or follow recognised guidance and your department’s advice, and justify your choice.

What p-value is considered significant?

The conventional threshold is .05, but it is a convention, not a law. Always report exact p-values (or p < .001) with effect sizes and confidence intervals, and interpret them in context.

What is Cronbach’s alpha?

A measure of the internal consistency of a multi-item scale. Values of about .70 or above are commonly treated as acceptable, though expectations vary by field and purpose.

What if my data are not normally distributed?

Check how serious the departure is. With larger samples, mild non-normality often matters less. Otherwise consider a non-parametric alternative or a justified transformation, and report your decision.

How do I report SPSS results in my dissertation?

Rebuild clean tables with key statistics, report test statistic, degrees of freedom, p-value and effect size in the text, and interpret the findings in relation to your hypotheses. Follow your department’s style, often APA 7.

Can I use free alternatives to SPSS?

Often yes, if your department allows. Jamovi, JASP and R are free and widely used. Check your guidance before switching.

Should I include SPSS output in my appendix?

If your department allows, yes. Appendices that include syntax, full output and a data dictionary support transparency. Make sure the data are anonymised.

Can someone help me with my SPSS analysis?

Yes, within your university’s rules. Supervisors, methods advisers and mentors can help you choose tests, run them and interpret results. Writers.ng offers guidance and feedback so that the analysis you submit is your own.

Continue through the dissertation process

Related guides in our cluster:

  1. Dissertation methodology chapter
  2. Qualitative vs quantitative dissertations
  3. Thematic analysis for your dissertation (NVivo guide)
  4. How to write a dissertation conclusion
  5. Dissertation proofreading checklist

Explore our wider support:

Stuck with your SPSS analysis?

Share your research question, your variables and where you are stuck. We will give honest, specific guidance on choosing tests, running them and interpreting the output, and leave the analysis in your hands.

Get guidance on your SPSS analysis