Common Data Analysis Mistakes
UAE Students Must Avoid

Khalifa University applies some of the most rigorous data analysis standards among UAE institutions, particularly at research master’s and PhD level. The most common rejection reasons in Chapter 4 submissions include:

• Tool mismatch: Using Excel for analysis that requires SPSS, R, or MATLAB at research level

• Missing assumption tests: Running regression or ANOVA without documenting normality, linearity, and multicollinearity checks

• Raw SPSS output presented as tables: SPSS screenshots with grey backgrounds and double-line borders are rejected on APA formatting grounds

• AI-generated interpretations: Turnitin Clarity flags are referred to the academic integrity committee at Khalifa University rather than being treated as a standard resubmission

If the rejection feedback is vague, request a detailed breakdown from your supervisor before beginning revisions. Addressing the wrong issue costs weeks.

The standard Turnitin similarity threshold applied at most UAE universities for postgraduate submissions is 15% or below for the overall similarity score, excluding the reference list and direct quotations. At PhD level, many supervisors expect the score to be closer to 10%.

However, the 2026 change that matters more for Chapter 4 specifically is not the similarity score — it is the AI writing detection score. Turnitin Clarity now produces a separate AI detection report. Most UAE universities, including UAEU and Khalifa University, are applying institutional thresholds to this score that, when exceeded, trigger an academic integrity referral independent of the text similarity result.

For data analysis chapters specifically, aim for an AI writing score below 10% and a similarity score below 15% before any supervisor submission. Always run the self-check independently before formal submission — discovering a flag after submission is significantly harder to resolve.

AUD applies APA 7th edition formatting to all dissertation tables and figures. The formatting process for SPSS output involves five steps that must be completed before the table is inserted into your Word document:

• Copy values only from the SPSS output viewer — never paste the formatted SPSS table as an image or object

• Rebuild in Word using a clean table with single-line borders at the top, below the header row, and at the bottom only — no vertical borders and no grey shading

• Add the table number above in plain text ( Table 1 ) followed by the bold title in title case on the next line, both left-aligned

• Standardise decimal places: two d.p. for means, standard deviations, and correlations; three d.p. for p-values with no leading zero (p = .032)

• Add a Note below the table if any abbreviations are used or if the data source requires attribution

AUD markers specifically check for raw SPSS output presentation — it is listed as a standard deduction item in the dissertation marking rubric for the presentation and rigour component.

No — Excel charts and SPSS tables inserted as images or values do not trigger Turnitin AI detection flags. Turnitin Clarity scans text content only, not images, embedded objects, or numerical data within tables.

The risk is entirely in the written text that surrounds those charts and tables. Specifically:

• The written interpretation below each table or figure

• The introductory paragraph for each analysis sub-section

• The summary statements at the close of each research question sub-section

If any of these text sections were generated using an AI tool and pasted without manual rewriting, Turnitin Clarity 2026 can identify the writing behaviour pattern as AI-assisted. The solution is to write all interpretation text manually, using your specific sample data and context — content that AI cannot generate without access to your actual findings.

One-way ANOVA requires three assumption tests to be completed and documented before results are presented in Chapter 4. These are mandatory at UAE research institutions including UAEU and Zayed University:

• Normality: Run the Shapiro-Wilk test (n < 50) or Kolmogorov-Smirnov test (larger samples) for each group being compared. If p > 0.05, normality is confirmed.

• Homogeneity of variance: Run Levene’s Test. If p > 0.05, equal variances are assumed. If p < 0.05, use the Welch ANOVA alternative and report it explicitly.

• Independence of observations: This is confirmed through the research design itself — document in your methodology that each respondent provided only one data point and that groups are mutually exclusive.

If normality fails, use the Kruskal-Wallis test as the non-parametric alternative to one-way ANOVA. Document the failed normality test, explain the decision to use the non-parametric alternative, and present Kruskal-Wallis results with the same rigour as you would ANOVA outputs.

Yes — non-significant results are entirely acceptable in a UAE university dissertation and must be reported in full. A p-value above 0.05 does not indicate a failed dissertation. It indicates that statistical significance was not established at the standard threshold for that particular relationship or comparison.

What is not acceptable is omitting or minimising non-significant results. Under the 2026 MoE academic integrity framework, selective reporting of only significant findings is classified as a research integrity violation. UAE university examiners at AUD, UAEU, and BUiD are specifically trained to look for missing tests or unexplained gaps in the results sequence.

The correct approach is to report the exact p-value, state that significance was not established at the 0.05 threshold, and reserve all interpretation of what this means — including limitations and alternative explanations — for Chapter 5. Non-significant findings often produce some of the most valuable discussion content in the dissertation.

In the UAE 6-chapter dissertation model, the boundary between Chapters 4 and 5 is one of the most commonly violated structural rules and one of the most consistent rejection reasons at BUiD and Zayed University.

• Chapter 4 — Data Analysis and Results: Presents what the data shows. Describes outputs, states statistical values, and answers research questions descriptively. No comparisons to literature, no theoretical interpretation, no implications statements.

• Chapter 5 — Discussion: Interprets what the findings mean. Connects results to the literature reviewed in Chapter 2, explains agreements and contradictions with prior research, identifies theoretical and practical implications, and acknowledges limitations.

A practical test: if any sentence in Chapter 4 references a prior study, compares a finding to existing theory, or uses words like “this suggests” or “this implies” in a conceptual sense, that sentence belongs in Chapter 5, not Chapter 4.

In most cases, a Chapter 4 rejection does not require new data collection. The most common rejection types are correctable using your existing cleaned dataset. The recovery depends on the category of rejection:

• Missing assumption tests: Re-run the analysis in SPSS with assumption tests enabled and document results. No new data required.

• Wrong tool: If your data is quantitative, re-run the analysis in SPSS using the same cleaned dataset. If qualitative, import transcripts into NVivo and rebuild the coding framework.

• APA formatting errors: Rebuild all tables in Word and reformat all figures. No analysis re-run required.

• Turnitin Clarity AI flag: Delete flagged interpretation text entirely and rewrite manually. Do not attempt to paraphrase AI-generated text — start from scratch with your actual outputs in front of you.

New data collection is only necessary if the supervisor has rejected the research instrument itself, identified a fundamental sampling design flaw, or determined that the existing dataset is too small to support the declared analysis method. These scenarios are less common and should be clarified in writing before any data collection decision is made.