Sampling Procedure
The primary focus of the 2013 VDHS was to provide reliable estimates of key population and health indicators, including fertility and mortality rates, both for the country as a whole, and separately for urban and rural areas (this is standard practice for a DHS). The survey used the sampling frame based on census enumeration areas, with population and household information from the 2009 Vanuatu Census of Population and Housing. The primary sampling units, comprising of 93 total enumeration areas (EA), were selected in each domain using systematic random sampling with probability proportional to the estimated number of households in the EA. Then in each selected EAs a total of 24 households were randomly selected with equal probability. It should be noted that DHS sampling was prepared by SPC.
It was not considered viable to generate results at an island division level for Vanuatu due to the expected small sample sizes at these fine geographical levels. However, it was considered worthwhile to split the rural population into two separate domains-Rural one covering households surrounding the urban area (easy access to Port-Vila and Luganville) and all households living in all administrative centers of all other provinces, and Rural 2 covering the other remaining rural Vanuatu population-because access to main health facilities for rural 1 population is more available compared to access to the Rural 2 population whom tend to have limited or no access to those health facilities.
The survey was designed to obtain completed interviews of 3,129 women aged 15-49. In addition, males aged 15 and over in every second household surveyed were interviewed. To take non-responses into account, 2,232 households countrywide were selected: 672 in the urban area and 1,560 in rural areas.
1. DOMAINS OF OBSERVATION - STRATIFICATION
Vanuatu 2012 DHS aims to provide results at National level first, but the distinction between rural and urban areas is expected as well.
The rural areas of Vanuatu cover a very large scope of households with different situation, especially in terms of access to health facilities. It was then considered to split the rural area of Vanuatu into 2 areas (Rural1 and Rural2)
There are therefore 3 domains of reporting for Vanuatu DHS 2013, which are based on splited Enumeration Area (EA) designed for the 2009 census:
Urban: people living in town - easy access to facilities (especially hospital)
.....Shefa:
..........Port Vila urban based on census definition
..........Rural areas around Vila urban (Mele, Pango, Erakor, Ifira)
...............Sanma:
...............Luganville urban based on census definition
Rural1: people living in areas with easy access to town, and people living in provincial centre with easy access to provincial facilities (especially provincial health centre).
..........Shefa: rest of Efate (out or Port-Vila): the ring road has changed the access to health facilities for the population living in North Efate, they have now a better access to Vila hospital
..........Sanma: around Luganville:
................On the West coast up to Port Orly
................On the East coast down to Tassiriki
................Aore island
..........Tafea: Lenakel
..........Ambae: Saratamata
..........Malampa: Lakatoro
..........Torba: Sola
Rural2: people living in area with no access to hospital or provincial health centre.
..........Shefa: all the islands out of Efate (Shepherd)
..........Sanma: Malo and far Santo rural
..........Tafea: Far Tanna rural and outer islands (Futuna, Aneytum, Aniwa, Erromango)
..........Penama: far Ambae rural and outer islands (Pentecost & Maewo)
..........Malampa: far Malakula, Ambrym, Paama
..........The rest of Vanuatu Rural will be classified as RURAL2 "Far rural"
There is no EA excluded from the selection. Each EA inhabited has a probability higher than 0 to be selected.
2. SAMPLE SIZE
The sampling unit is the household, and the sampling was bassed on the Vanuatu 2009 census. DHS sample is based on 2 major constraints, where in each domain, there was a need to interview at least:
..........1000 women 15-49
..........500 men 15+
The women constraint was the most important, as more individuals wrre needed from a smaller age group. That was why the sample design was based on the distribution of women 15-49 in Vanuatu.
The sample size in each domain was determined according to the average number of women 15-49 per households:
........Women aged 11-45 years old at the time of 2009 census (11/2009) who will be 15-49 years old on 11/2013 (period of DHS field work).
In order to get robust estimates, 1000 women were needed in each of the 3 areas. These 1000 women are then allocated within each stratum according to the number of women in each stratum.
Based on the number of households, we could derive the average number of women 15-49 per households in 2013
At this stage, we could automatically derive the number of households to select per stratum (number of women to interview / average women 15-49 per household)
3. SELECTION OF PSU (splited EAs)
The sampling strategy is a 2 stages sample:
........Selection of Primary Sampling Unit (splited EAs)
........Within all the selected splited EAs, selection of households.
Primary Sampling Unit is the first stage of the selection. In the case of Vanuatu, PSU are the splited Enumeration Area used for the 2009 census. According to the number of household expected in each domain and the workload of 24 households per EA (number of household one team can survey in 1 week), we can come out with the following number of EAs to select:
2232 households have to be interviewed following this distribution per stratum. This sample size allows the interview of at least 1000 women per domain, which is the main constraint. In order to achieve this sample size, field teams have to replace any refusals or non contacted households.
All the inhabited splited EAs of Vanuatu have a probability of selection which is proportional to its size (in terms of number of households). This is the probability of selection of each splited EA:
4. NUMBER OF FEMALES 15-49 AND MALES 15+ EXPECTED IN THE SAMPLE
As expected, at least 1000 women were obtained in each domain.
Only 500 males 15+ are required per domain. If all the males of all selected households are interviewed, there will be more than twice as many men as needed for the survey purpose.
In the field we need the men's sample half the size of the women's sample. That way field staff can interview males 15+ from every second household selected (12 households).
5. SELECTION OF HOUSEHOLDS (household listing)
· The second stage of the sampling plan is the selection of 24 households within the selected EAs. All the households within the EA have the same probability of selection. As we expect 24 households per EA, the probability of selection for each household is: 24/total no. of hhlds in EA
In order to get 24 households, and to prevent any replacement (refusal, absence...), it is preferable to select 6 extra households. Now we have to select:
...........24 households to contact and interview in priority- listA
...........6 households in case of replacement - listB
==> Total 30 households
The 24 households selected in listA are the one to contact first. If for any reason the interview cannot be conducted in some of these households, interviewers have to replace with households listed in listB (6 extra households)
The latest household listing comes from the 2009 census, which is 4 years old now. Since the last census, the household listing have not been updated. Field teams have to achieve this task in the field before they start the interviews. The household listing is the first exercise each team has to complete. They have to list all the occupied dwelling within the EA. Any vacant house should not be recorded in the household listing.
Once the household listing is achieved (household counted from 1 to N), supervisors have to do the household selection, 30 households (24 listA and 6 listB)
In order to select 30 households out of the updated household list, supervisor need:
...........At least 30 households within the EA (if in the updated listing, less than 24 households have been identified, some extra households have to be selected in the nearest EA),
...........A sampling step 'St'
...................St = Total no. of HHS in the EA/30 ---- keep only the integer,
...................A starting point Sp = the supervisor choose a number between 1 and the sampling step 1 <= Sp <= St
From the starting point (Sp), add the sampling step (St) separately to get the serial number (1 to N) of the other 29 selected households.
Within these 30 selected households, 4/5 are listed in listA and 1/5 in listB. It means 4 out of 5 selected households are listA and the fifth is listB.
In the following example, 74 households are listed in the household listing.
...........The sampling step St = 74/30 = 2.46 ==> 2 is the integer
...........The starting point Sp=(1,2) ==> 1 or 2 are possible, 2 is selected by the supervisor
...........The selection of the 30 households is shown in the following table:
In case of more than 6 replacements are required (listB too short), field staff can choose any households within the EA.
...........In each selected EA, a survey team will be in charge of 24 households. One week is the time period required to interview all males aged 15 and more, and all females 15 - 49 in each household (= one round of field operation)
...........At the end of each round, each team has completed 24 households