Sampling Procedure
A sample of 2700 smokers was targeted from 135 enumeration areas. A total of 20 smokers was targeted in each EA. The specific steps for the selection of the EAs and smokers included the following stages:
Step 1: Sampling of EAs
From the full list of all 4,000 plus EAs in the Gambia, disaggregated by strata, EAs were randomly selected from within the specified strata/district based on the number of EAs
Step 2: Sampling of smokers/households in each EA
Random Number Table (RNT) sampling was used to select eligible households when the number of eligible households in an EA was more than 20 households. The team in an EA randomly selected a start point to administer the household questionnaire (up to 20 administered in each EA). Within each household, if a smoker was present and willing and eligible to be interviewed, then a second, individual questionnaire was administered. All households with eligible smokers in the EA were relisted. A sampling interval was determined for systematic random sampling method
3 starting numbers are randomly selected and the enumeration starting with the first random number selected in the first round of 20 households with smokers. If, after the first round of 20 households were interviewed, there were fewer than 20 questionnaires, the second random start number was used to select another set of 20 households to interview. This was repeated until at least 20 households were interviewed in the EA. All households were interviewed in EAs with fewer than 20 eligible households. If more than one smoker is present in the household, interviewers randomly selected one smoker to be interviewed
Weighting
The inverse of three probabilities of selection were combined to design the weights:
1. The probability of selecting the EA - Number of EA selected/ Total number of EAs
2. The probability of selecting the 20 block of HHs - 20 households selected/Average number of households in an EA
Note that to estimate the average number of households in an EA the total number of households in an LGA were first divided by the total EAs in an LGA. The numbers were between 29 and 90 households per EA which gives an average of about 53 households per EA. Gambia's 2003 census report estimates that the average size of an EA is about 65 households or 500 people. Gambia's 2013 census found 217,000 households in total. Using the average of 53, the Project estimated 212,000 households.
3. The probability of selecting the smoking HH - smoker/Number of smokers in household
Note that all households with a smoker were selected for interviews, i.e. Probability=1. However, within each household, only one smoker was interviewed, based on the probabilities calculated
To calculate the base weights we these probabilities were multiplied by their inverses. The final weights were adjusted according to response rates at household and individual smoker levels.
