Quick Answer
Probability sampling designs give every population element a known chance of selection: simple random, stratified, systematic, cluster and multi-stage. Non-probability designs — convenience, judgment, quota, snowball — cannot support statistical inference. Design choice trades cost against precision.
Visual Summary
- Simple random — lottery; every unit equal chance.
- Stratified — homogeneous layers sampled proportionally.
- Cluster — whole natural groups sampled at once.
- Quota — non-random; interviewer fills fixed quotas.
Probability Designs
Simple random sampling: every unit has an equal chance (lottery or random numbers); it needs a full sampling frame and works best for homogeneous populations. Stratified sampling: divide the population into internally homogeneous strata (income bands, regions), sample each — guarantees representation of every stratum and shrinks sampling error because between-strata variance is removed. Systematic sampling: pick every k-th unit after a random start (k = N/n); efficient with ordered frames but dangerous with periodic lists. Cluster sampling: sample naturally occurring groups (villages, schools) and survey all or some members — cheaper field logistics, larger sampling error. Multi-stage sampling chains these: states, then districts, then households.
Non-Probability Designs
Convenience sampling takes whoever is available — fastest, most biased. Judgment sampling relies on expert choice. Quota sampling fills fixed demographic quotas non-randomly — the only common design with no probability basis in selection. Snowball sampling recruits through referrals — essential for hard-to-reach populations (informal workers, migrants) at the cost of unknown selection probabilities.
Choosing a Design
The decision trades frame availability, cost and precision: a complete frame favours simple random; known strata favour stratification; dispersed populations favour clusters; no frame at all forces non-probability designs — whose findings carry selection-bias caveats rather than sampling-error estimates.
Exam Angle
NET passages pair a research scenario with the right design — match the selection mechanism to the name. Practise in the UGC NET Commerce mock.
FAQ
Which design minimises sampling error for a given size?
Stratified — removing between-strata variance shrinks error relative to simple random.
Why is quota sampling non-probability?
Interviewer choice fills the quotas; selection probabilities stay unknown.
Exam angle
How exams use this: Every lane here has a timed mock that mirrors its real marking scheme — reading is step one, recall is the exam. Test it under exam conditions →
Related reading
- Sufi and Bhakti Movements Compared: Saints, Orders and Spread | Medieval India One-Pager for Exams
- Indian Society Part 9: Migration and Remittances — A Nation on the Move
Quick revision
- Simple random: — lottery; every unit equal chance.
- Stratified: — homogeneous layers sampled proportionally.
- Cluster: — whole natural groups sampled at once.
- Quota: — non-random; interviewer fills fixed quotas.
- Sufi and Bhakti Movements Compared: Saints, Orders and Spread | Medieval India One-Pager for Exams
- Indian Society Part 9: Migration and Remittances — A Nation on the Move
Have a doubt on this topic?









