Sampling
Sampling Procedure
At the national level, we have a two-stage stratified non-self weighted of equal size for both “Intervention” and “Control” groups. For each of the two study groups the sample size has been determined as of 1200 households3. The determination of the sample size was guided by the previous experience of EICV surveys in which the smallest sample size for an analysis domain was found to be in the vicinity of 500 households. In addition, due to time and resource constraints it would be difficult to implement a sample of a bigger size. The original sampling scheme is described subsequently; nonetheless there has been some deviation from the initial design as explained in due course.
The designed sample is 120 households in each selected sector whether intervention or control. Due to some non-respondent cases the implemented sample is somewhat less than the designed one. Non-response occurred primarily for different causes such as “failure to locate the sample household in the field”; households are “Not-at home” during data collection period. “Refusals” were found to be inexistent.
Response Rate
A designed sample of 1200 households in which the implemented sample was 1175households, which gives the response rate of 97.9 %.
Weighting
The basic weight for each sample household is equal to the inverse of its probability of selection (calculated by multiplying the probabilities at each sampling stage). Since all survey data has been processed by computer, it was easy to attach a weight to each sample household record in the computer files, and the tabulation programs can weight the data automatically. The sampling probabilities at each stage of selection were maintained in an Excel spreadsheet so that the overall probability and corresponding weight ware calculated for each sample sector. (See report, Annex B for details).
