Nursing Dissertation Guide

How to Analyse Nursing Dissertation Data Using SPSS: A Step-by-Step Guide

Learn how to analyse nursing dissertation data using SPSS, from data preparation and descriptive statistics to hypothesis testing, interpretation and reporting.

Written by Moses Updated August 25, 2026 15 min read
How to Analyse Nursing Dissertation Data Using SPSS: A Step-by-Step Guide

Analyse nursing dissertation data using SPSS with a structured process that connects every statistical decision to your research question. SPSS can organise data, calculate descriptive statistics, test differences, examine relationships and create tables. However, the software cannot decide whether your variables, sample, assumptions or study design justify a particular test. That judgement remains the researcher’s responsibility.

This practical guide is the broad SPSS hub for the site. It explains the full workflow from a raw dataset to a defensible results chapter. For deeper guidance on one stage, use the focused guides on choosing statistical tests, interpreting SPSS output, writing the SPSS results chapter, analysing Likert-scale data and running regression analysis.

Key Takeaways

  • Start with the research question rather than the SPSS menu.
  • Keep an untouched copy of the original dataset.
  • Label, code and clean every variable before formal analysis.
  • Choose tests according to variable type, study design and assumptions.
  • Report effect sizes and confidence intervals alongside p-values.
  • Use reformatted tables instead of pasting raw SPSS output.
  • Save syntax or an analysis log so the work can be reproduced.

How to Analyse Nursing Dissertation Data Using SPSS

Begin by translating every research question into an analytical task. Identify the outcome variable, predictor or grouping variable, level of measurement and type of question. A descriptive question asks what the sample or data look like. A comparative question examines differences between groups or time points. A relational question investigates whether variables move together. A predictive question asks whether one or more factors explain an outcome.

For example, a question asking whether a nurse-led education programme improves medication-adherence scores contains a continuous outcome and a comparison. When the same patients are measured before and after the programme, the observations are related. When two separate groups are compared, they are independent. This distinction affects the selected test.

Create an analysis plan before running procedures. For each objective, record the variables, coding, descriptive statistics, proposed inferential test, assumptions, significance threshold, effect-size measure and intended table. This plan reduces random testing and makes the methodology easier to defend.

Prepare the Dataset Before SPSS Analysis

In a standard SPSS file, each row normally represents one participant or observation, while each column represents one variable. Use short and meaningful variable names without spaces. Add full variable labels so another person can understand the dataset. For categorical variables, define value labels such as 0 = No and 1 = Yes.

Set the measurement level carefully. Nominal variables are categories without an inherent order, such as ward, diagnosis or treatment group. Ordinal variables have a meaningful order, such as satisfaction levels. Scale variables are continuous or treated as continuous, such as age, blood pressure, length of stay or a validated total score.

Keep the original file unchanged. Save a working copy and use clear version names. Document every recode, calculated variable, case exclusion and correction. SPSS syntax is valuable because it records the commands used and makes the work easier to repeat after supervisor feedback.

Clean Nursing Dissertation Data in SPSS

Data cleaning should occur before hypothesis testing. Run frequencies for categorical variables and descriptive summaries for continuous variables. Look for values outside the permitted range, duplicate records, inconsistent category codes, unexpected blanks and impossible dates.

For example, a pain score of 77 on a scale ranging from 0 to 10 is probably an entry error. An age of 4 would also be impossible in a study restricted to adults. Check questionable values against the source record rather than deleting them automatically.

Confirm that units are consistent. Mixing kilograms and pounds, centimetres and inches, or mmol/L and mg/dL can invalidate an analysis. Check whether questionnaire items require reverse scoring before calculating a total score. Keep a cleaning log that states what was changed, when it was changed and why.

SPSS distinguishes user-defined missing values from system-missing values, so missing codes such as 99 or 999 need to be declared rather than treated as genuine observations (IBM, n.d.-d).

Handle Missing Data Transparently

Missing data should never be silently converted to zero. First calculate the amount of missing information for each variable. Then examine whether missingness is concentrated among particular groups, time points or questionnaire items. Systematic missingness can bias the findings.

SPSS procedures may exclude cases listwise or pairwise. Listwise deletion removes a case when any required value is missing. Pairwise deletion uses the available values for each calculation, which means the sample size can differ across results. IBM’s Missing Value Analysis documentation describes these approaches and shows how the procedure can examine patterns and extent of missing data (IBM, n.d.-b).

Complex imputation should not be used merely because it sounds advanced. The strategy must fit the amount and pattern of missingness, sample size, design and available methodological expertise. IBM notes that its Missing Value Analysis procedure can describe missingness and estimate statistics under several approaches, while multiple imputation is generally preferable to simple single imputation when imputation is justified (IBM, n.d.-b).

Produce Descriptive Statistics

Descriptive statistics explain who participated and what the main variables looked like before inferential testing. For categorical data, report frequencies and percentages. For approximately symmetrical continuous data, the mean and standard deviation may be appropriate. For skewed data or data affected by influential outliers, the median and interquartile range may provide a more representative summary.

Describe the sample before presenting tests of difference or association. Relevant characteristics may include age, gender, clinical setting, diagnosis, professional experience, treatment group and baseline score. Avoid presenting every variable merely because SPSS generated it. Each statistic should help describe the sample or answer a research question.

IBM describes the Explore procedure as a tool for data screening, outlier identification, distribution review and assumption checking (IBM, n.d.-a).

Check Statistical Assumptions

Parametric procedures rely on assumptions that differ by test. These may include independent observations, approximate normality, homogeneity of variance, linearity and absence of influential outliers. Check the relevant assumptions before interpreting results, not only after an unexpected p-value appears.

Do not judge normality from one formal test alone. Consider histograms, Q-Q plots, skewness, kurtosis, sample size and the nature of the variable. In large samples, a formal normality test can identify a small departure that has little practical importance. In very small samples, graphical review and cautious interpretation become especially important.

For independent-group comparisons, Levene’s test helps assess equality of variances. When equal variances cannot be assumed, report the corrected result rather than automatically using the first row of output. For regression, review linearity, residuals, influential observations and multicollinearity.

Choose the Correct SPSS Statistical Test

Research task Common parametric option Possible alternative Key issue
Compare two independent groups Independent-samples t-test Mann-Whitney U Groups must be unrelated
Compare two related measurements Paired-samples t-test Wilcoxon signed-rank Same participants or matched pairs
Compare three or more groups One-way ANOVA Kruskal-Wallis Post-hoc tests may be required
Relate two continuous variables Pearson correlation Spearman correlation Check the relationship pattern
Associate categorical variables Chi-square test Exact test where needed Review expected cell counts
Predict a continuous outcome Linear regression Depends on the design Check residuals and multicollinearity

This table is a starting point rather than an automatic decision rule. Variable measurement, sample structure, distribution, design and assumptions still determine whether the procedure is defensible. For a more detailed decision tree, use the dedicated Statistical Tests for Nursing Dissertations guide.

Run and Record the SPSS Analysis

Use the Analyse menu or syntax to run the selected procedure. For every test, record the variables included, valid sample, assumption checks, test statistic, degrees of freedom, exact p-value, effect size and confidence interval where relevant.

Avoid repeatedly testing different variable combinations until a significant result appears. This increases the risk of false-positive findings and makes the analytical pathway difficult to defend. Exploratory analysis can still be useful, but it should be labelled as exploratory rather than presented as if it was planned from the beginning.

Save the output file and syntax. Name versions clearly and retain a brief analysis log. Reproducibility becomes especially important when the dissertation is revised after supervisor feedback.

Interpret P-Values, Effect Sizes and Confidence Intervals

A p-value does not show the probability that the research hypothesis is true. The American Statistical Association explains that a p-value indicates how incompatible the data are with a specified statistical model and warns against treating a threshold alone as evidence of scientific importance (Wasserstein & Lazar, 2016).

Likewise, a non-significant result does not prove that no difference exists. The study may have limited power, an imprecise measurement tool or a small sample. Interpret the p-value with the design, sample size, estimated effect and confidence interval.

Effect sizes describe the magnitude of a difference or relationship, while confidence intervals communicate precision and uncertainty. Biomedical reporting guidance recommends presenting estimates and uncertainty rather than reducing results to significance declarations alone (Lang & Altman, 2015).

For a table-by-table approach to reading SPSS results, continue to Nursing SPSS Output Interpretation.

Build Composite Questionnaire Scores Correctly

Many nursing dissertations use multi-item questionnaires to measure knowledge, confidence, attitudes, wellbeing or quality of life. Before calculating a total score, check the scoring instructions provided by the instrument developer. Some items may require reverse coding, and some tools permit a total score only when a minimum number of items is complete.

Use Transform and Compute Variable to create the score, but verify the formula carefully. IBM notes that ordinary arithmetic and statistical functions can handle missing values differently; for example, a simple arithmetic expression can become missing when one component is missing, whereas a MEAN function can be configured around available values (IBM, n.d.-c). The correct approach must still follow the validated instrument’s scoring rule.

Reliability analysis may be relevant when evaluating internal consistency. Cronbach’s alpha is commonly reported, but it is not proof that a scale is valid or unidimensional. Avoid deleting items solely to increase alpha without a theoretical and methodological justification. The Likert-scale SPSS guide covers coding, reverse scoring and scale construction in more detail.

Plan Sample Size and Statistical Power

A small sample can produce unstable estimates, wide confidence intervals and insufficient power to detect a meaningful difference. A large sample can make a very small effect statistically significant. Therefore, sample size should be considered before data collection and revisited during interpretation.

The calculation depends on the main analysis, expected effect size, significance level, desired power and number of predictors or groups. When using an existing dataset, explain the available sample and acknowledge limitations rather than presenting a post-hoc calculation as if it determined recruitment.

Attrition and incomplete questionnaires can reduce the final analytical sample. Report how many people were approached, enrolled, excluded and analysed where the design requires it. This helps readers judge selection bias and understand why the denominator differs between tables.

Create Clear Tables and Figures

Tables should answer a question efficiently. Give each table a descriptive title, define abbreviations and state the valid sample where necessary. Use a consistent number of decimal places and avoid false precision. Percentages should include the denominator when missing data cause the base to change.

Figures are useful for distributions, group comparisons and trends, but decorative charts can distract from the evidence. Choose a graph that matches the variable type and avoid three-dimensional effects. Check that axes use defensible values and labels remain readable.

Refer to every table or figure in the narrative. Summarise the main pattern rather than repeating every cell. The reader should understand why the visual is included and how it relates to the research question.

Write the Results Chapter Clearly

Organise the results around the research questions or objectives. Begin with participant flow, missing data and sample characteristics. Then present descriptive and inferential findings in a logical sequence. Use concise narrative text to guide the reader to tables instead of repeating every number.

A useful reporting sequence is to state what was tested, name the procedure, report relevant descriptive statistics, provide the test statistic and p-value, include the effect size and confidence interval, and state the immediate result. SAMPL guidance similarly emphasises transparent reporting of statistical methods and results, including sufficient numerical information for readers to evaluate the findings (Lang & Altman, 2015).

Do not paste raw SPSS tables into the main dissertation. Reformat relevant information into clear academic tables. Full output may be included in an appendix when the institution requires it, but an appendix should not replace proper reporting in the chapter. The dedicated Nursing SPSS Results Chapter guide focuses on that write-up stage.

When Regression Needs Its Own Analysis Plan

Regression is not simply the next step after correlation. The outcome type, predictor selection, coding, sample size, assumptions and model purpose all need to be justified. A model should be built from the nursing question rather than from every variable available in the dataset.

For linear and logistic regression, coefficient interpretation, odds ratios, diagnostics and model reporting, use the focused Nursing Regression Analysis in SPSS guide.

Document Decisions and Protect Research Integrity

Statistical analysis should leave an audit trail. Record the dataset version, exclusions, recoding decisions, calculated variables, assumptions, procedures and output files. When supervisor feedback leads to a changed analysis, explain what changed and why. This separates exploratory work from the final analytical plan.

Remove direct identifiers before importing data into SPSS and store files according to the approved data-management plan. Do not delete, alter or selectively report observations merely because they weaken the preferred conclusion. A defensible nursing dissertation reports unexpected, non-significant and uncertain findings honestly.

Before submission, compare the results with the approved methodology. Every promised test should appear or have a clear explanation for its omission. Every reported procedure should also have a methodological justification.

Common SPSS Mistakes in Nursing Dissertations

  • Choosing a test because it is familiar rather than appropriate.
  • Treating ordinal categories as continuous without justification.
  • Failing to declare missing-value codes.
  • Ignoring influential outliers or impossible values.
  • Reporting percentages without the denominator.
  • Reporting only p-values and omitting effect sizes.
  • Claiming causation from a cross-sectional association.
  • Describing a non-significant result as proof of no effect.
  • Copying every SPSS output table into the chapter.
  • Changing the planned analysis after viewing results without explanation.

When Specialist Data Analysis Support Helps

Project-specific support may help when the dataset contains complicated coding, repeated measurements, several outcomes, missing data, scale construction or regression models. It may also be useful when supervisor feedback says the analysis is unclear, incomplete or inconsistent with the methodology.

Our specialist nursing data analysis support can review the question, variables, dataset, assumptions, selected tests and results chapter. For broader help, explore our nursing dissertation services, review our matching and quality standards, or request a final dissertation editing review.

Final Checklist to Analyse Nursing Dissertation Data Using SPSS

  • Every analysis answers a stated question or objective.
  • Variables are labelled, coded and measured consistently.
  • Missing data, errors and outliers have been reviewed.
  • Descriptive statistics suit the variable distributions.
  • Assumptions have been checked and documented.
  • The test matches the variables and study design.
  • Exact p-values, effect sizes and confidence intervals are reported.
  • Tables are reformatted rather than copied directly from SPSS.
  • Conclusions do not exceed what the design can support.
  • Syntax or an analysis log makes the work reproducible.

Frequently Asked Questions

Can SPSS choose the correct statistical test?

No. SPSS performs the selected procedure. The researcher must decide whether it matches the question, variables, design and assumptions.

Should every nursing dissertation use inferential statistics?

No. Some projects are descriptive, qualitative, evidence reviews or service evaluations. Inferential testing should be used only when it fits the design and objectives.

Is a statistically significant result clinically important?

Not necessarily. Statistical significance alone does not establish the magnitude, clinical relevance or practical importance of a result (Wasserstein & Lazar, 2016).

Should raw SPSS output appear in the dissertation?

Usually, only selected and reformatted findings belong in the main chapter. Full output may be placed in an appendix when required.

What should I do when assumptions are not met?

Confirm that coding is correct, assess the seriousness of the violation and consider a corrected, robust or non-parametric approach where justified. Explain the decision clearly.

Can someone review analysis I have already completed?

Yes. Provide the research questions, methodology, anonymised dataset, syntax or output, draft results and supervisor feedback.

Related Nursing Guides

Conclusion

To analyse nursing dissertation data using SPSS effectively, begin with the research question, prepare a transparent analysis plan, clean the dataset and select procedures that match the variables and design. Strong analysis goes beyond producing a significant p-value. It reports descriptive findings, assumptions, effect sizes, confidence intervals and limitations so that readers can evaluate the evidence.

References

Related articles

Browse all