Sampling
Sampling Procedure
Rwanda is administratively divided into four provinces (Northern Province, Southern Province, Eastern Province and Western Province) plus Kigali City and a total of 30 districts. Districts are further divided in sectors and cells. To facilitate comparison with existing studies, the CFSVA 2015 was designed to provide statistically representative
and precise information at the district level. In addition, it was decided to include both urban and rural households
and not to exclude the capital province Kigali. The sampling frame was organized according to the 30 districts.
Subsequently, a two-stage cluster sample procedure was applied.
In the first stage, 25 villages per district were randomly selected with probability to be selected proportional to the
population size. In the second stage, ten households in each of the 25 villages in the 30 provinces were selected for
participation in the survey. A systematic random sampling technique was chosen for this stage. The team leader, together with the village head, listed all households in the village. Based on this list, a systematic random sample
was utilized to pick ten households to be interviewed and three reserve households should any of the first ten
households be missing at the time of the interview or not agree to participate. Households were eligible for
participation in the assessment if living in one of the selected villages at the time of the interviews.Thus, ten households, from 25 villages, from 30 provinces were chosen to participate in the survey, amounting up
to 7,500 households.
The sample size was not designed to produce precise estimates for malnutrition prevalence at district level. The primary goal of collecting the nutrition data was to analyse the link between food security and nutrition. Also, information from key informants was collected through a structured questionnaire but the sample was not designed to be statistically representative for villages in Rwanda; the information from the community questionnaire was therefore used for contextual information only.
Weighting
Taking into consideration the two-stage cluster sampling methodology described above, adjustment weights were
computed to provide results representative at country level. The household probability of being selected in the
sample is equal to the product of a household’s probability of being selected in a village by the probability of the
village of being sampled. The inverse of this probability is the design weight. The design weight was adjusted for
the expected and actual number of households in the surveyed villages and was used in the complex sample
calculations. The design weight was divided by the product of the total number of households in the population
divided by the number of sampled households. The resulting weight was used in all non-complex sample analyses.
Documentation in PDF