Data analysis is where many dissertations stall. The data is collected, the methodology chapter is approved, and then the questions start. Which test answers my research question? Do I need to check assumptions? Why does SPSS give me three different tables? How do I write this up without interpreting it?
This guide works through those questions in the order you will meet them: matching each research question to a method, preparing and checking the data, running the analysis, and reporting it in a results chapter (usually Chapter 4) that your committee will accept.
It covers quantitative and qualitative analysis, and it explains what an analyst can and cannot do for you.
How to choose your tests
Start from your research questions and hypotheses, identify the type of each variable, and choose the test that fits that combination (for example, an independent t-test for comparing two groups on a continuous outcome). Clean and screen the data, check the test's assumptions, run the analysis, and report each result in APA format: the test, the statistic, degrees of freedom, p-value, effect size and a one-sentence plain statement. Save interpretation for the discussion chapter.
Step 1: Build an Analysis Plan from Your Research Questions
Your methodology chapter should already name the analysis. If it does not, or if the data has turned out differently from what you expected, write a one-page analysis plan before running anything. For each research question, record:
- the hypothesis (null and alternative), if quantitative;
- the variables involved and their level of measurement (nominal, ordinal, interval or ratio);
- which variable is the outcome and which are predictors or groups;
- the test or method you will use, and its assumptions;
- what you will do if an assumption is violated.
This plan keeps you from running every test in the menu and reporting whichever gives a significant result, which committees recognize and penalize.
Matching the Question to the Test
| Your question | Variables | Parametric test | Non-parametric alternative |
|---|---|---|---|
| Do two groups differ? | 2 independent groups, continuous outcome | Independent samples t-test | Mann-Whitney U |
| Did scores change over time? | Same people measured twice | Paired samples t-test | Wilcoxon signed-rank |
| Do three or more groups differ? | 3+ groups, continuous outcome | One-way ANOVA | Kruskal-Wallis |
| Are two variables related? | Two continuous variables | Pearson correlation | Spearman correlation |
| Are two categories associated? | Two categorical variables | Chi-square test of independence (Fisher's exact for small samples) | |
| What predicts an outcome? | Continuous outcome, several predictors | Multiple linear regression | Robust or ordinal regression |
| What predicts a yes/no outcome? | Binary outcome | Binary logistic regression | |
| Do several items measure one construct? | Questionnaire items | Cronbach's alpha, exploratory or confirmatory factor analysis | |
Likert-scale data is a common source of argument. Single Likert items are ordinal and suit non-parametric tests; composite scales built from several items are usually treated as continuous. State your reasoning in the methodology and follow your program's convention.
Step 2: Prepare and Screen the Data
Most errors in results chapters start before the analysis. Work through this list and record what you did, because the results chapter should report it briefly.
| Check | What to look for | Typical action |
|---|---|---|
| Data entry | Impossible values (age 250, a 7 on a 5-point scale) | Correct from source or set to missing |
| Missing data | How much, and whether it is random | Report the amount; use listwise deletion, multiple imputation or another method justified in your methodology |
| Reverse-coded items | Negatively worded questionnaire items | Recode before computing scale scores |
| Outliers | Extreme values on boxplots or z-scores beyond about 3.29 | Check, then keep, transform or remove with a stated rule |
| Normality | Histograms, Q-Q plots, Shapiro-Wilk (with caution in large samples) | Proceed, transform, or use a non-parametric test |
| Reliability | Cronbach's alpha for each scale | Report alpha; values around .70 or above are usually considered acceptable |
Step 3: Run the Analysis and Check Assumptions
Each test has assumptions, and your committee will expect you to have checked them. For example, an independent t-test assumes independent observations, an approximately normal outcome in each group, and similar variances (Levene's test).
Multiple regression adds linearity, independence of errors, homoscedasticity and the absence of severe multicollinearity (variance inflation factors well below 10). When an assumption fails, report it and use the alternative you named in your plan, such as Welch's t-test or a non-parametric test.
Software
| Software | Best for | Note |
|---|---|---|
| SPSS | Most social science, education, nursing and business dissertations | Menu-driven; save the syntax so the analysis can be reproduced |
| R or Python | Complex models, large datasets, reproducible scripts | Free; supervisors in quantitative fields often prefer them |
| Stata | Economics, epidemiology, panel data | Strong for regression and survey weights |
| NVivo, ATLAS.ti, MAXQDA | Qualitative coding of interviews and documents | Software organizes coding; it does not do the analysis for you |
Step 4: Report Results in APA Style
Report each result the same way: the test, the descriptive statistics, the test statistic with degrees of freedom, the exact p-value, an effect size with its interpretation, and one plain sentence saying what was found.
Example: An independent samples t-test showed that nurses on 12-hour shifts reported higher burnout (M = 3.41, SD = 0.62) than nurses on 8-hour shifts (M = 3.02, SD = 0.71), t(186) = 3.98, p < .001, d = 0.58, a medium effect. Hypothesis 1 was supported.
Use tables for anything with more than three or four numbers, such as descriptive statistics, correlation matrices and regression coefficients, and refer to each table in the text without repeating every value. Report non-significant results in the same detail as significant ones.
Effect Sizes to Report
| Test | Effect size | Small / medium / large (Cohen's guidelines) |
|---|---|---|
| t-test | Cohen's d | 0.2 / 0.5 / 0.8 |
| ANOVA | Partial eta squared | .01 / .06 / .14 |
| Correlation | r | .10 / .30 / .50 |
| Regression | R squared, standardized beta | Interpret in context |
| Chi-square | Cramér's V or phi | Depends on degrees of freedom |
Worked Example: From Research Question to Reported Result
Here is how one research question travels through the whole process in a typical education dissertation.
Research question: Is there a relationship between weekly hours of paid employment and academic self-efficacy among community college students, after controlling for age and enrollment status?
Variables: outcome, academic self-efficacy (scale score, continuous); predictor, hours of paid work per week (continuous); covariates, age (continuous) and enrollment status (full or part-time, dummy coded).
Test: hierarchical multiple regression, with covariates entered in step 1 and work hours in step 2.
Checks: scale reliability (alpha = .86), linearity and homoscedasticity from residual plots, VIFs below 2, no influential cases (Cook's distance below 1), 14 cases removed for missing data (5.1%).
Reported result: After controlling for age and enrollment status, hours of paid work explained an additional 6% of the variance in academic self-efficacy, ΔR² = .06, F(1, 255) = 17.42, p < .001. More hours of work were associated with lower self-efficacy, β = −.25, p < .001.
Notice the language: "associated with", not "caused". A correlational design cannot support a causal claim, and committees notice when the results chapter overreaches.
Structuring Chapter 4
- Introduction: one paragraph restating the purpose and research questions.
- Data screening and sample: response rate, missing data, exclusions, and participant characteristics in a demographics table.
- Reliability and assumptions: scale reliabilities and assumption checks, briefly.
- Results by research question: one sub-section per question or hypothesis, in the same order as Chapter 1.
- Summary: a short paragraph or table listing each hypothesis as supported or not supported.
The rule committees enforce most often: Chapter 4 reports, Chapter 5 interprets. Phrases such as "this suggests", "this may be because" and comparisons with other studies belong in the discussion.
Mistakes Committees Flag Most Often
Methodologists and committee members see the same problems in results chapters year after year. Checking your draft against this list before submission saves a round of revisions.
| Mistake | Why it is a problem | Fix |
|---|---|---|
| Running many tests and reporting only the significant ones | Inflates false positives and looks like p-hacking | Report every planned test; correct for multiple comparisons where appropriate (for example Bonferroni or Holm) |
| Writing "p = .000" | A p-value is never exactly zero | Report p < .001 |
| No effect sizes | Significance alone says nothing about practical importance | Report an effect size for every test |
| Treating "not significant" as "no difference" | Absence of evidence is not evidence of absence, especially in small samples | Say the difference was not statistically significant and discuss power in Chapter 5 |
| Pasting raw SPSS tables | Default output is not in APA format and includes irrelevant rows | Rebuild tables in APA style with only the values you discuss |
| Changing the analysis without saying so | Committees compare Chapter 4 against the approved plan | Explain any deviation from the proposal and why it was necessary |
| Interpreting in Chapter 4 | Blurs results and discussion | Move explanations and comparisons with the literature to Chapter 5 |
Qualitative Analysis
Qualitative dissertations need just as much rigor, but the steps differ. Thematic analysis following Braun and Clarke's six phases is the most common approach: familiarization, initial coding, generating themes, reviewing themes, defining and naming themes, and writing up.
Other approaches include interpretative phenomenological analysis, grounded theory and content analysis; each has its own procedures, so use the one named in your methodology.
A qualitative findings chapter is organized by theme, with each theme defined, explained and supported by participant quotations that show its range.
Include a table of themes and sub-themes, and describe how you established trustworthiness: an audit trail, member checking, reflexive notes, or a second coder. For a deeper treatment, see our guide to qualitative research methods.
How Long Each Stage Takes
| Stage | Clean survey data, 2-4 tests | Complex models or messy data | Qualitative, 10-20 interviews |
|---|---|---|---|
| Analysis plan | 1 day | 2-3 days | 1-2 days |
| Cleaning and screening | 1 day | 3-5 days | Transcription checks, 2-4 days |
| Analysis | 1-2 days | 4-7 days | Coding and theme development, 1-2 weeks |
| Write-up and tables | 2-3 days | 4-6 days | 4-7 days |
Build in time for your chair's feedback. Results chapters are commonly revised at least once, often because the committee wants an additional check or a different presentation of a table.
What an Analyst Can Do, and What Stays Yours
| Can do | Must stay yours |
|---|---|
| Write or refine the analysis plan | The data, collected as approved by your ethics board |
| Clean, code and screen your dataset | Decisions your committee has already approved |
| Run the tests and check assumptions | Understanding of what was done, so you can defend it |
| Produce APA tables and the results write-up | The interpretation in your discussion chapter, informed by your field knowledge |
| Explain the output in plain language |
Results are always produced from your real data. No reputable analyst will fabricate or adjust data to reach significance, and doing so is research misconduct.
What to Send
- Your dataset (SPSS .sav, Excel, CSV, or transcripts for qualitative work) with a codebook or questionnaire
- Research questions and hypotheses
- Your methodology chapter, or the approved proposal
- Your university's results chapter template or formatting rules
- Any feedback from your chair or methodologist
You receive the results write-up, the tables, and the output or syntax files, so you can show exactly what was run.
Get Your Results Chapter Done Properly
Send your dataset, research questions and methodology. A statistician runs the right tests on your data and reports them in APA style, with output files included.
Start My Data AnalysisFree revisions · Full refund if late · Written from scratch for your order
Frequently Asked Questions
It depends on your research question and the type of variables. Comparing two groups on a continuous outcome suggests a t-test; three or more groups suggests ANOVA; relationships suggest correlation or regression; categorical associations suggest chi-square. Your analysis plan should justify the choice.
Report them fully and honestly. Non-significant results are still findings. Discuss possible reasons, such as sample size or measurement, in Chapter 5, and never change the analysis just to reach significance.
Yes. Send the dataset and the questionnaire. Scale scores, reliability, descriptive statistics and the tests that answer your research questions can be run and reported in APA style.
Yes. You receive the output and syntax or script files along with the write-up, so the analysis can be checked and reproduced.
No. Chapter 4 reports what the data shows. Chapter 5 interprets the findings, compares them with the literature and states implications.
A straightforward quantitative analysis with a clean dataset can take a few days. Complex models, messy data or qualitative coding of many interviews take longer, often one to three weeks.