Approach | Pros | Cons |
---|---|---|
Multi-stage cluster sampling | ||
Census frame with manual cluster selection |
Calculate weights and CIs Low cost of 1st-stage cluster selection Does not require skills beyond survey statistics |
Outdated 1st-stage frame High cost of 2nd-stage household enumeration |
Gridded population frame with GridSample.org |
Calculate weights and CIs Low cost of 1st-stage cluster selection Requires few skills beyond survey statistics |
Outdated 1st-stage frame High cost of 2nd-stage household enumeration |
Gridded population frame with RTI Geo-Sampling |
Calculate weights and CIs Requires few skills beyond survey statistics |
Outdated 1st-stage frame High cost of 1st-stage cluster selection High cost of 2nd-stage household enumeration |
Census frame with EPI design |
Low cost of 1st-stage cluster selection Low cost of 2nd-stage Does not require skills beyond survey statistics |
Outdated 1st-stage frame Cannot calculate weights and CIs |
Spatial sampling | ||
No stratification |
Sample not representative of the population Requires GIS skills | |
Stratification on structure density |
Updated frame Calculate weights and CIs |
Requires GIS and other skills beyond survey statistics High cost of structure enumeration (or similar count of buildings) |
Simple random sample of structures | ||
Enumerate all structures in OpenStreetMap, Google Earth, or GIS software |
Updated frame Calculate weights and CIs |
High cost of structure enumeration Requires GIS skills |
Epicentre Geo-Sampler |
Updated frame Calculate weights and CIs Low cost of structure selection (no enumeration) Requires few skills beyond survey statistics |