Questionnaire Study Guide

📖 Core Concepts Questionnaire – a research tool consisting of a set of written questions used to collect data from respondents. Variable‑based questionnaire – measures separate variables (e.g., behaviours, facts). Scale‑based questionnaire – aggregates items into a composite score that taps a latent trait (e.g., attitude). Structured vs. Unstructured – Structured: identical item order for everyone; Unstructured: free‑form text, no fixed format. Closed‑ended vs. Open‑ended – Closed: respondent chooses from given options; Open: respondent generates their own answer, later coded. Response scales – dichotomous (yes/no), nominal‑polytomous (unordered categories), ordinal‑polytomous (ordered categories), bounded continuous (numeric range). Multi‑item scale – ≥3 items measuring the same construct, usually on a 5‑ to 7‑point Likert‑type rating scale. Reliability & validity – internal consistency, test‑retest, content, construct, and criterion validity ensure a scale measures what it should. 📌 Must Remember Standardized answers → easy coding & analysis. Exhaustive & mutually exclusive options are required for closed‑ended items. One construct per item – avoid double‑barreled questions. Positive wording preferred; avoid negatives & double negatives. Logical flow: screening → warm‑up → main items → sensitive/difficult → demographics. Multi‑item scales need ≥3–5 items, balanced verbal anchors, and a consistent rating scale. Reliability types: internal (item‑item consistency) and test‑retest (stability over time). Validity hierarchy: content → construct → criterion. TRAPD translation steps – Translation, Review, Adjudication, Pretest, Documentation. 🔄 Key Processes Designing a questionnaire Define constructs → write one‑construct items → choose wording (positive, clear). Decide questionnaire type (variable‑based vs. scale‑based; structured vs. mixed). Select response format (dichotomous, nominal, ordinal, continuous). Order items using logical and sensitivity progression. Building a multi‑item scale Draft ≥5 items per construct. Apply a consistent 5‑ to 7‑point Likert rating (e.g., 1 = Strongly disagree, 7 = Strongly agree). Conduct pilot, run factor analysis → keep items loading on intended factor, drop weak items. Compute internal reliability (e.g., Cronbach’s α). Translation (TRAPD) Translation: multiple translators produce drafts. Review: compare drafts, resolve discrepancies. Adjudication: expert panel decides final wording. Pretest: administer to native speakers, check comprehension. Documentation: record decisions for future reference. 🔍 Key Comparisons Variable‑based vs. Scale‑based – measures separate facts vs. aggregates into a latent score. Structured vs. Unstructured – fixed order & options vs. free‑form text. Open‑ended vs. Closed‑ended – respondent creates answer vs. selects from list. Dichotomous vs. Nominal‑polytomous – 2 mutually exclusive options vs. >2 unordered categories. Ordinal‑polytomous vs. Bounded continuous – ordered categories (e.g., “agree” levels) vs. numeric interval (e.g., 0–100). ⚠️ Common Misunderstandings “Open‑ended = no coding needed.” – responses must still be coded for analysis. “More items always improve reliability.” – irrelevant or poorly worded items can lower internal consistency. “Translation is word‑for‑word.” – cultural adaptation is essential; literal translation may change meaning. “All respondents understand every item.” – assume literacy; pilot test for clarity. 🧠 Mental Models / Intuition “One‑construct, one‑item” – picture each question as a single spotlight shining on one trait. “Scale as a ruler” – a multi‑item scale is like a ruler: each tick (item) adds precision to the measurement of an invisible length (latent trait). “Flow as a story arc” – start easy (intro), build tension (core), climax (sensitive), wrap up (demographics). 🚩 Exceptions & Edge Cases Low‑literacy populations – may require pictorial or interviewer‑administered modes. Highly sensitive topics – sometimes better to place early if “response mode” cannot be guaranteed later. Online surveys – risk of extremely low response rates; may need incentives or follow‑ups. 📍 When to Use Which Variable‑based questionnaire → when you need distinct, factual data points (behaviour, demographics). Scale‑based questionnaire → when measuring attitudes, socioeconomic status, or other latent constructs. Structured format → large samples, need for comparability. Unstructured/pictorial → exploratory work, low‑literacy or visual‑learning groups. Dichotomous → simple presence/absence decisions, screening. Ordinal‑polytomous → attitudinal intensity (e.g., Likert). Bounded continuous → precise numeric input (e.g., temperature, rating 0–100). 👀 Patterns to Recognize Mutual exclusivity & exhaustiveness → every closed‑ended set should cover all possible answers without overlap. Balanced anchors → Likert scales should have equal intervals (e.g., “Strongly disagree” ↔ “Strongly agree”). Skip‑logic branches → look for “If = yes, go to Q5; else go to Q6.” Item‑total correlation → low correlation flags a problematic item in a scale. 🗂️ Exam Traps “All closed‑ended questions are dichotomous.” – false; closed‑ended includes nominal and ordinal polytomous scales. “A multi‑item scale must have exactly 5 items.” – not required; ≥3 is sufficient if reliability is high. “Translation guarantees equivalence.” – without pretesting, cultural nuance can be lost, leading to measurement error. “Higher response rate always means better data.” – if non‑respondents differ systematically, bias remains. “Skip logic can be ignored in analysis.” – it changes the sample composition for downstream items; must be accounted for.