LAWS6032 Chap.6 Surveys, Sampling and Quantitative Analysis
Surveys, Sampling and Quantitative Analysis
Survey estimates are jointly created by the sampling frame, questionnaire and response process. This chapter compares probability and non-probability sampling, question formats, coding, missing data and response strategies. Quantitative examples keep numerators and denominators visible, distinguish observed respondent proportions from population prevalence, and explain what weighting can and cannot repair.
What this chapter covers
- 01
Target populations and sampling frames
- 02
Probability, stratified, cluster and systematic samples
- 03
Purposive, quota, convenience and snowball recruitment
- 04
Structured questions and balanced response options
- 05
Coverage, nonresponse and measurement error
- 06
Coding plans and missing data
- 07
Response rates, proportions and subgroup comparisons
- 08
Weighting for known population differences
Calculating and labelling a survey result
- +2Response rate = 208 divided by 320 multiplied by 100 = 65%.
- +2Observed concern proportion = 104 divided by 208 multiplied by 100 = 50% among respondents.
- +2Do not extend the 50% to all eligible users without addressing whether response relates to concern and comparing known characteristics.
Key terms
- Target population
- The full set of people or units to which the research question and intended inference refer.
- Sampling frame
- The operational list or process through which members of the target population can enter selection.
- Probability sample
- A design in which selection probabilities are known or calculable for units in the frame.
- Nonresponse bias
- Distortion arising when response relates to the study variables after accounting for the design.
- Response rate
- The proportion of eligible invited or sampled cases that provide a usable response under a stated rule.
- Survey weight
- A multiplier adjusting each responding case's contribution for selection or known population differences.
- Likert item
- A structured response item asking degree of agreement, frequency or another ordered judgement on a defined scale.
Surveys, Sampling and Quantitative Analysis FAQ
Why does a larger sample not guarantee representativeness?
Sample size affects random precision, while representativeness depends on frame coverage, selection and response. A very large convenience sample can precisely describe its respondents yet still misrepresent people who lacked access, declined or were never eligible to appear.
When is non-probability sampling appropriate?
It can be appropriate for exploratory work, specialised knowledge, qualitative inquiry or hard-to-reach populations. The proposal should explain the selection logic and keep the inference within the recruited cases rather than attaching probability-based population language without support.
What makes a survey question double-barrelled?
A double-barrelled item asks about two constructs while permitting one answer, such as whether police were responsive and fair. Respondents may hold different views on each part. Split the item so every response has an unambiguous interpretation.
What can weighting repair?
Weighting can align the sample with known population margins or unequal selection probabilities. It cannot recover people entirely absent from the frame, correct invalid questions, or guarantee that respondents and nonrespondents are equivalent on unmeasured characteristics.
How should I code missing survey answers?
Distinguish structural not applicable responses, explicit refusal, do not know answers and accidental omissions. Predefine codes before collection and use the eligible item base as the denominator. Collapsing these categories can distort response quality, subgroup comparison and the meaning of the resulting percentage.
Assessment move
Practise drawing the target population, frame, invited sample, respondents and analysed cases as nested sets. At each transition, write who can be lost and whether that loss may relate to the outcome. Recalculate response rates, proportions and simple weights with every step shown. Always label the population after the percentage.
Rewrite leading, double-barrelled, vague and unbalanced questions into neutral single-construct items with an appropriate timeframe. Compare probability and non-probability designs by the inference they support rather than assuming one is always superior. For every weighting exercise, inspect the reason weights differ and state the variance or residual-bias cost.
End survey interpretations with separate sentences for the observed respondent result and the additional assumption needed to extend it. Pilot questions by asking someone to explain how they chose an answer; hesitation often reveals ambiguous timeframes or categories. Create a missing-data codebook before collection so refusal, not applicable and skipped items are not collapsed.
For weighted results, calculate population share divided by sample share, then describe why a large weight increases the influence of relatively few cases. Compare weighted and unweighted estimates as a diagnostic. Keep the symbols p, π and σ attached to their roles so sample observations are not casually renamed as population parameters.
Working through Surveys, Sampling and Quantitative Analysis in LAWS6032? Sia is AskSia’s AI Law tutor — ask any LAWS6032 Surveys, Sampling and Quantitative Analysis question and get a clear, step-by-step explanation grounded in how LAWS6032 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
