Rwanda - Rwanda Post Enumeration Survey 2012
Reference ID | rwa-nisr-rpes-2012-v1 |
Year | 2012 |
Country | Rwanda |
Producer(s) | National Institute of Statistics of Rwanda - Ministry of Finance and Economic Planning |
Sponsor(s) | Government of Rwanda - GoR - Financial support One UN - One UN - Financial assistance United Nations Population Fund - UNFPA - Financial assistance United Nations Children's Fund - UNICEF - Financial assistance UKaid - - Financial ass |
Metadata |
![]() |
Created on
May 21, 2015
Last modified
May 21, 2015
Page views
44193
Sampling
Sampling Procedure
Conceptually, the PES involves two samples, named the "population" P sample and the "Enumeration" E sample. The P sample consists of the PES sample of segments (Enumeration Areas, EA's) drawn from the same target population, but independently from the census, for the purpose of estimating census omissions when compared to Census records. The E sample is drawn from the cases already enumerated in the Census, but selected for independent re-enumeration for the purpose of estimating census erroneous inclusions when compared to the original Census records. Although the E sample may be separate from the p sample, in practice it is made to overlap completely with the P sample to reduce costs and improve the precision of the estimates. The E sample then consists of the same EA's selected for the PES. A two-way match is conducted between the P sample and the E sample to identify both the omissions and erroneous inclusions. The matching also produces estimate of matched population required in the dual-system estimator of the true population.
The Enumeration Areas, as defined in the mapping operation implemented prior to the 2012 Census, is the Primary Sampling Unit (PSU), while the private household is the Ultimate Sampling Unit (USU). As all Six households included in the sample EA's are included in the sample with certainty, the selection probability of a household is exactly equivalent to the selection probability of the corresponding EA. The EA's list created during the mapping stage constitutes the frame of the EA's. Beside the geographic specification the frame includes estimates of the number of households and the number of the population in each EA. The total number of EA's in Rwanda is 16716 with an average size of 128.6 households each. The size dispersion of EA's is nearly moderate, the standard deviation is about 35.2 households and the coefficient of variation is 27.3 percent. About 80.1 percent of EA's are between 90 and 180 households, while only 1.9 percent of EA's are as small as 60 households or less, there exists about 1 percent of EA's sized 210 or more households.
The normal choice of stratifying variable is the type of residence place (urban, semi-urban, and rural), previous PES surveys in Rwanda (1991 and 2002) exhibit disparity of net coverage error rate between urban and rural. In addition to such explicit stratification of the sampling frame, an implicit stratification based on geographic proximity is also introduced during the sampling selection operation.
The literature review of previous Post Enumeration Surveys carried out in Rwanda (1991, 2002) has revealed that the adopted sample size was 120 EA's for both indicated surveys. As such it was deemed appropriate and logical to maintain this size of the sample for the present PES. Nonetheless, the sample size was independently calculated based on anticipated coverage rate of 97%, deff =2, confidence coefficient of 95%, relative error margin within 10%, and average size of EA of about 128.6 households and about 600 persons and the number of strata is 3, the resulting sample is about 124 EA's which is only 4 EA's greater than the adopted sample size for the present PES. In case of higher coverage rate the relative error margin would be slightly greater than the assumed level of 10%.
The sample was allocated over the strata in such a way that: Urban sample is 40 EA's and Semi-urban sample is 35 EA's.
Deviations from Sample Design
The standard deviation is about 35.2 households.
Response Rate
The response rate of listed households exceeds 99 percent at the national level.
Weighting
To obtain unbiased estimates from the PES data it has been necessary to apply appropriate weights to the sample data based on the probabilities of selection. It was also important to calculate measures of sampling variability for Census coverage and content estimates.
In order to avoid producing biased sample estimates, it was necessary to multiply the data by a sampling weight, or expansion factor. The basic weight for each sample household member was equal to the inverse of his/her probability of selection. As indicated before, since all households and household members were included in the PES Sample with certainty, the selection probability of a certain EA was exactly equivalent to the selection probability of a certain household and a household member within this EA.