Sampling

The sample for the Kyrgyzstan Multiple Indicator Cluster Survey was designed to provide representative estimates of MICS indicators at the national level, in urban and rural areas, as well as for eight regions: Batken, Jalalabad, Issyk Kul, Naryn, Osh, Talas, Chui regions, and Bishkek. The urban and rural areas of each region were used as strata, where the sample design was made in two stages.

The principal objective of the sample design was to provide current and reliable estimates on a set of indicators covering the four major areas of the World Fit for Children declaration, including promoting healthy lives; providing quality education; protecting against abuse, exploitation and violence; and combating HIV/AIDS.  The population covered by the 2006 MICS is defined as the universe of all women aged 15-49 and all children aged under 5.  A sample of households was selected and all women aged 15-49 identified as usual residents of these households were interviewed.  In addition, the mother or the caretaker of all children aged under 5 who were usual residents of the household were also interviewed about the child.

Four hundred clusters, or Census-1999 Enumeration Areas (CEA), were selected with a probability proportional to the population size in the first stage. For rural areas, populated settlements were used as Primary Sampling Units (PSUs). For urban areas, internal territorial-administrative units were used as PSUs. For each enumeration area, a household listing was updated and used as a sample framework for the second selection stage. Later, households with an equal probability were
selected, according to the up-dated lists of addresses.

In defining the cluster size, a high rate of intra-cluster correlation of different indicators was taken into account. This required clusters of small size, as well as consideration of the effective use of interviewers’ time, requiring a minimization of movement from one settlement to another. As a compromise between data accuracy and the efficient use of limited time and funding, a cluster size was determined to consist of 13 households.

Thus, a total sample volume consisted of 5,200 households. Given that a sample is not self-weighting, and that sample size by strata is approximately equal, sample weights were used for reporting national level results.

The sampling procedures are more fully described in the sampling design document and the sampling appendix of the final report.

No major deviations from the original sample design were made.  All sample enumeration areas were accessed and successfully interviewed with good response rates.

During the course of the survey, all 400 PSUs selected at the first sampling stage were visited. A list of household addresses was made for those PSUs. Out of 5,200 sample households, 5,199 were found to be occupied (Table ??.1). Out of these populated households, 5,179 were successfully interviewed, yielding a household response rate of 99.6%. In all regions except for Naryn, the interviewers managed to carry out interviews in all selected households.

In the interviewed households 7,043 women (aged 15-49) were identified. Of these women, 6,973 were successfully interviewed, which corresponds to a response rate of 99.0%. Additionally, the household sample accounted for 3,000children under five years of age, and 2,987 questionnaires were completed on these, for a response rate of 99.6.

Sample weights were calculated for each of the datafiles.  

Sample weights for the household data were computed as the inverse of the probability of selection of the household, computed at the sampling domain level (urban/rural within each region).  The household weights were adjusted for non-response at the domain level, and were then normalized by a constant factor so that the total weighted number of households equals the total unweighted number of households.  The household weight variable is called HHWEIGHT and is used with the HH data and the HL data.

Sample weights for the women's data used the un-normalized household weights, adjusted for non-response for the women's questionnaire, and were then normalized by a constant factor so that the total weighted number of women's cases equals the total unweighted number of women's cases.

Sample weights for the children's data followed the same approach as the women's and used the un-normalized household weights, adjusted for non-response for the children's questionnaire, and were then normalized by a constant factor so that the total weighted number of children's cases equals the total unweighted number of children's cases.