Demographic and Health Survey 2005

SAMPLE DESIGN

Creation of the 2005 CDHS sample was based on the objective of collecting a nationally representative sample of completed interviews with women and men between the ages of 15 and 49. To achieve a balance between the ability to provide estimates for all 24 provinces in the country and limiting the sample size, 19 sampling domains were defined, 14 of which correspond to individual provinces and 5 of which correspond to grouped provinces.

• Fourteen individual provinces: Banteay Mean Chey, Kampong Cham, Kampong Chhnang, Kampong Speu, Kampong Thom, Kandal, Kratie, Phnom Penh, Prey Veng, Pursat, Siem Reap, Svay Rieng, Takeo, and Otdar Mean Chey;
• Five groups of provinces: Battambang and Krong Pailin, Kampot and Krong Kep, Krong Preah Sihanouk and Kaoh Kong, Preah Vihear and Steung Treng, Mondol Kiri, and Rattanak Kiri.

The sample of households was allocated to the sampling domains in such a way that estimates of indicators can be produced with known precision for each of the 19 sampling domains, for all of Cambodia combined, and separately for urban and rural areas of the country.

The sampling frame used for 2005 CDHS is the complete list of all villages enumerated in the 1998 Cambodia General Population Census (GPC) plus 166 villages which were not enumerated during the 1998 GPC, provided by the National Institute of Statistics (NIS). It includes the entire country and consists of 13,505 villages. The GPC also created maps that delimited the boundaries of every village. Of the total villages, 1,312 villages are designated as urban and 12,193 villages are designated as rural, with an average household size of 161 households per village.

The survey is based on a stratified sample selected in two stages. Stratification was achieved by separating every reporting domain into urban and rural areas. Thus the 19 domains were stratified into a total of 38 sampling strata. Samples were selected independently in every stratum, by a two stage selection. Implicit stratifications were achieved at each of the lower geographical or administrative levels by sorting the sampling frame according to the geographical/administrative order and by using a probability proportional to size selection at the first stage of selection.

In the first stage, 557 villages were selected with probability proportional to village size. Village size is the number of households residing in the village. Some of the largest villages were further divided into enumeration areas (EA). Thus, the 557 CDHS clusters are either a village or an EA. A listing of all the households was carried out in each of the 557 selected villages during the months of February-April 2005. Listing teams also drew fresh maps delineating village boundaries and identifying all households. These maps and lists were used by field teams during data collection.

The household listings provided the frame from which the selection of household was drawn in the second stage. To ensure a sample size large enough to calculate reliable estimates for all the desired study domains, it was necessary to control the total number of households drawn. This was done by selecting 24 households in every urban EA, and 28 households in every rural EA. The resulting oversampling of small areas and urban areas is corrected by applying sampling weights to the data, which ensures the validity of the sample for all 38 strata (urban/rural, and 19 domains).

All women age 15-49 years who were either usual residents of the selected households or visitors present in the household on the night before the survey were eligible to be interviewed. In addition, in a subsample of every second household selected for the survey, all men age 15-49 were eligible to be interviewed (if they were either usual residents of the selected households or visitors present in the household on the night before the survey). The minimum sample size is larger for women than men because complex indicators (such as total fertility and infant and child mortality rates) require larger sample sizes to achieve sampling errors of reasonable size, and these data come from interviews with women.

In the 50 percent subsample, all men and women eligible for the individual interview were also eligible for HIV testing. In addition, in this subsample of households all women eligible for interview and all children under the age of five were eligible for anemia testing. These same women and children were also eligible for height and weight measurement to determine their nutritional status. Women in this same subsample were also eligible to be interviewed with the cause of death module, applicable to women with a child born since January 2002.

The 50 percent subsample not eligible for the man interview was further divided into half, resulting in one-quarter subsamples. In one-quarter subsample all women age 15-49 were eligible for the woman's status module in addition to the main interview. In this same one-quarter subsample, one woman per household was eligible for the domestic violence module. In the other one-quarter subsample, women were not eligible for the woman's status module, nor the domestic violence module.

NOTE: See detailed description of the sample design in APPENDIX A of tthe survey report.