ATLASResearch
methods
Quantitative + Mixed methods/ Sampling

Cluster sampling

Select groups before selecting people or records

Cluster sampling first selects groups such as schools, bank branches or geographic areas. A one-stage design observes all units in selected clusters; multistage designs sample within them. Observations within a cluster are often correlated, so effective information can be far smaller than the raw number of individuals. Selection probabilities and cluster structure must carry through into analysis.

WHEN IT FITS

Use when populations are geographically or organisationally grouped and a full individual-level frame is unavailable or dispersed data collection would be prohibitively expensive.

Strengths

  • Reduces fieldwork and travel costs
  • Allows sampling through organisational or geographic frames

Limitations

  • Within-cluster similarity reduces precision
  • Too few sampled clusters weaken variance estimation

Know the boundary

A large number of people from a few schools does not equal the information in the same number sampled independently.

USED ACROSS
Banking & financeBusiness & MBAPsychologyComputer scienceEducation
CONNECTED METHODS