Survey design for UAE student researchers — the working notes

Surveys are the default quantitative method for UAE undergraduate and master’s dissertations. They’re popular because they look approachable — write some questions, distribute on WhatsApp, get responses, analyse in SPSS. The reality is that surveys are one of the easiest methodologies to do badly, because every weak choice (vague question, leading prompt, poor scale design, biased sample) compounds into data that can’t actually answer the research question.

Here is how we sequence survey design at the studio when we work with dissertation clients.

Step 1: Align the survey to the research question, not the other way around

The most common dissertation failure mode is writing the survey before the research question is fully crystallised. The student knows roughly what they want to study, drafts a survey that “feels right”, distributes it, and then discovers in the analysis phase that the data can’t answer the actual question they care about.

The fix is to write the research question first — in a single sentence — and then design each survey item to contribute directly to answering it. If a candidate question doesn’t map to your research question, cut it.

A useful check: for each survey question, ask “what statistic from this question goes into my findings chapter, and which sentence of my conclusion does it inform?” If you can’t answer both, the question doesn’t belong.

Step 2: Pick a scale type per question

The default for UAE dissertation surveys is a 5-point Likert scale (Strongly Disagree → Strongly Agree). It’s familiar to respondents, easy to analyse, and works for most attitudinal questions.

But the default isn’t always right. Three other scale types deserve consideration:

• 7-point Likert — useful when you need finer discrimination (e.g., comparing two groups whose attitudes might differ subtly). Statistically more sensitive, but slightly harder for respondents.
• Semantic differential (bipolar adjectives — Innovative ←→ Conservative with 7 points) — better for measuring perceptions of objects (brands, organisations).
• Frequency or magnitude scales — for behavioural questions (Never / Rarely / Sometimes / Often / Always) or amounts (Less than 10% / 10–25% / 25–50% / >50%).

Mix scale types within one survey, but be aware that mixing changes the analysis options. Mean scores across a Likert scale and a frequency scale aren’t comparable.

Step 3: Avoid the four classic question design errors

• Double-barrelled questions — I find this product reliable and good value. The respondent might find it reliable but poor value. Split into two questions.
• Leading prompts — To what extent do you agree that this poorly-designed product… The phrasing telegraphs the desired answer. Reword neutrally.
• Negative phrasing mixed with agreement scales — I do not find this product unreliable on a Strongly Disagree → Strongly Agree scale. Confuses respondents.
• Jargon or assumed knowledge — Rate the company’s commitment to ESG materiality assessment. If your respondents don’t all know the term, you’re measuring familiarity with jargon, not the construct.

Step 4: Decide sample size before distribution

For UAE undergraduate dissertations, the realistic sample-size target is 100–200 respondents. For master’s, 150–300. Anything significantly below makes inferential statistics unreliable; anything significantly above is fine but requires more time than you have.

A more rigorous approach is to do a power calculation in G*Power (free software). For a typical correlation analysis with medium effect size (r = 0.3), power = 0.80, and alpha = 0.05, you need roughly 84 respondents. For regression with three predictors and the same parameters, you need closer to 100. Run the calculation appropriate to your planned analysis.

Step 5: Plan the distribution channels honestly

UAE student researchers typically distribute through three channels: their own WhatsApp and Instagram networks; LinkedIn or professional contacts; and snowball sampling through friends and colleagues. Each channel introduces its own sampling bias — your friends are demographically similar to you; LinkedIn skews professional; snowball produces a network-cluster sample.

There is no perfectly representative way to sample for a UAE undergraduate dissertation. There is a way to be honest about the limitations in your methodology chapter. Name your channels, acknowledge the sampling bias, and frame your findings as inference about the sampled population rather than the full population.

Step 6: Pilot the survey on 5–10 respondents before mass distribution

Pilot testing catches the question-design errors that survived your own review. Send the survey to 5–10 respondents in the target population, ask them to flag anything ambiguous, and then revise.

The single most useful pilot question to add at the end is: Was any question unclear? If so, which one and why? You’ll be surprised what ambiguities you missed.

Step 7: Ethics approval — sequence this with item finalization

UAE universities all require ethics approval for human-subjects research. Apply early. The submission usually requires the final survey instrument, the participant information sheet, and the consent form. Most institutions turn approvals around in 2–4 weeks, but some take longer in busy windows.

Don’t distribute before approval lands. Pre-approval distribution invalidates the data for your dissertation.

Where The Essay Atelier fits

For dissertation projects involving primary survey research, our methodology design support includes survey instrument design, ethics approval drafting, and (where the client wants it) SPSS or R analysis of the collected data. The writer-editor pair walks through the steps above before any drafting starts.

If you’re at the stage of designing a survey and want a second opinion on whether your items align with your research question before you distribute, send the editors the question list. The review is short, the data quality protected is substantial.