A stratified two-stage cluster sampling methodology was applied: In the first stage, Enumeration Areas (polygons) were selected within each stratum. In a second step, households were randomly sampled in each polygon. Results are statistically representative at the stratum level, permitting comparative analysis between the target groups in each stratum. Specifically:
IDP sample & limitations:
- The sample was distributed across neighbourhoods (admin level 3) proportionately to the estimated IDP population size (as per IOM/DTM estimates from Round 22). Within each neighbourhood (admin level 3), polygons were created to capture the areas with high density of IDPs. The samples were further distributed across these polygons proportionately to the population density of the polygons. An interval for the interviews was calculated and every Xth household was interviewed. If the selected household was not an IDP, the interviewer continued to the next until the target sample per polygon was reached.
- For the IDP camps, no polygons were needed, the sample was distributed across the Sites proportionately to the population size and the households were randomly interviewed within the Site based on a set interval.
- Representativity: The IDP samples in Pemba are representative of areas with higher presence of IDPs; the IDP samples in Metuge and Mecufi are representative of neighborhoods hosting IDPs; and the IDP sample in the Sites are representative of the sites across Metuge and Mecufi.
Non-displaced sample & limitations:
- Enumeration areas (EA) were randomly sampled by INE in strata 1 and 2 of Pemba. In each EA, 30 households were randomly interviewed.
- Representativity: The non-displaced sample is representative of the local population at strata level in Pemba. For Strata 3 in Mecufi and Metuge: the non-displaced samples were distributed across the same polygons created for the IDPs (areas hosting IDPs). The sample is thus representative of the hosting areas and not of the district.
The IDP samples were calculated based on the DTM round 22 population estimates for each of the 4 strata and ensuring a 5% margin of error, 95% confidence. Because of the two-stage sampling design, a design effect of 1.2 was also added.
The non-displaced samples were calculated based on a standard maximum population size, 5% margin of error, 95% confidence and again adding the design effect of 1.2.
