Survey ID Number

AGO_2011_MIS_v01_M

Title

Malaria Indicator Survey 2011

Sampling Procedure

OBJECTIVES OF THE SAMPLING DESIGN

(1) The 2011 AMIS survey was designed to determine reliable malaria prevalence estimates among children under age 5 at the various domains of interest (when feasible) and mortality estimates for children under age 5.

(2) The major domains to be distinguished in the tabulation of key indicators are:

- Angola at the national level

- The majority of indicators for each of the four domains defined for Angola and classified as the following regions:

1) Hyperendemic region, high malaria prevalence

2) Mesoendemic Stable region, medium malaria prevalence

3) Mesoendemic Unstable region, medium malaria prevalence, though prevalence is affected by the amount of rain

4) Luanda province

- Urban and rural areas of Angola (each as a separate domain)

- Any contiguous group of provinces with an adequate sample size of at least 1,500 households

(3) The primary objective of the 2011 AMIS is to provide estimates with acceptable precision for important population indicators associated with each domain, such as:

a. Ownership and use of mosquito bednets.

b. Practices to treat malaria among children under age 5 and the use of specific antimalarial drugs

c. Prevalence of malaria and anemia among children age 6-59 months

d. Knowledge, attitudes, and practices regarding malaria in the general population

SAMPLE FRAME

Administratively, Angola is divided into 18 provinces, which can be grouped into eight subregions depending on how they share some common factors.2 In turn, each province is subdivided into municipalities (164 in total), and each municipality is divided into communes (532 in total). Each commune is classified as either urban or rural. In addition to these administrative units, in preparation for the last population census, each urban commune was subdivided into segments named census sections (CSs) that were equivalent to enumeration areas. The National Statistical Institute (INE) had been preparing cartographic materials, including a count of rooms and dwellings, for each CS in the urban areas. This material became an appropriate sampling frame for the 2011 AMIS. However, INE does not have updated cartographic material for the rural areas. To compensate for this lack, INE uses its regional offices to collect a list of villages, with estimated populations in each village, for most of the rural communes,. To develop the sample frame for the 2011 AMIS, the list of CSs was used for the urban communes and the list of villages was used for the rural communes.

STRATIFICATION

The communes were grouped by major region, by rural or urban location, by sub-region, and by province as a way to identify homogeneous sampling units. In addition, within each urban commune, several CSs were grouped, taking advantage of the existing neighborhoods (sub-districts) for stratification of the sample.

SAMPLE SIZE

The following table includes different scenarios used to select a sample size in a populationbased survey. In the absence of domains, the numbers are valid for the entire population; however, if analyses are expected for more than one domain, then the numbers should be interpreted as required for each domain.

SAMPLE ALLOCATION

The clusters for the implementation of the 2011 AMIS are defined on the basis of census sections (CSs) for urban communes and on the basis of villages for rural communes. The 240 clusters considered for the 2011 AMIS were equally allocated at 60 clusters in each domain. The target for the 2011 AMIS was to select about 8,800 households. Therefore, the sample take is on average 36 selected households per cluster (i.e., 8,800/240). Clusters are distributed as 96 in the urban areas and 144 in the rural areas.

Under the final sample allocation, it is expected that each of the four major malaria regions in Angola will provide a minimum of about 2,200 completed women interviews, 2,100 children under age 5, and 2,000 births in the last five years. Neither the distribution of the 240 clusters among major regions nor the distribution of households in the sample is proportional to the estimated population distribution. This is due to the disproportional number of CSs among major regions. As a result, the sample for the 2011 AMIS is not a selfweighted household sample. Therefore, the 2011 AMIS sample is unbalanced for residence areas and regions and will require the design of a final weighting adjustment procedure to provide representative estimates for all the study domains.

SAMPLE SELECTION

The sample for the 2011 AMIS was selected using a stratified three-stage cluster design consisting of 240 clusters, with 96 in urban areas and 144 in rural areas. In each urban or rural area in a given region, clusters are selected systematically with probability proportional to size.

The sampling procedures are fully described in Appendix A of " Angola Malaria Indicator Survey 2011 - Final Report" pp.43-48.

Estimates of Sampling Error

The sample of respondents selected in the 2011 AMIS is only one of many samples that could have been selected from the same population, using the same sample design and expected size. Each of these samples would yield results that differ somewhat from the results of the actual sample selected. The variability that would be observed between all possible samples constitutes the sampling error. Although the degree of variability is not known exactly, it can be estimated from the sample actually selected.

A sampling error is usually measured in terms of the standard error (SE). The standard error for a mean, percentage, difference, or any other statistic calculated from the data in the sample can be defined as the square root of the variance, which is a measure of the variation in all possible samples. For example, for any given statistic calculated from the sample, the value of that statistic will fall within a range of plus or minus two times the standard error of that statistic in 95 percent of all possible samples of identical size and design.

If the sample of households had been selected as a simple random sample, it would have been possible to use straightforward formulas for calculating sampling errors and the limits for the confidence intervals. However, as has been mentioned, the 2011 AMIS sample is the result of a complex, multi-stage stratified design, and, consequently, it was necessary to use more complex formulas that take the effects of stratification and clustering into consideration. It was possible to calculate the sampling errors for the 2011 AMIS using a computer program known as Module for Sampling Errors, included in the computer package ISSA (Integrated System for Survey Analysis). This program processes percentages or medians as a ratio estimate r = y/x where both the numerator y and the denominator x are random variables.

Sampling errors for the 2011 AMIS are calculated for selected variables considered to be of primary interest. The results are presented in this appendix for the country as a whole, for urban and rural areas, and for each endemic region. For each variable, the tables present the value of the statistic (R), its standard error (SE), the number of unweighted (N) and weighted (WN) cases, the design effect (DEFT), the relative standard error (SE/R), and the 95 percent confidence limits, i.e., the values R+2SE and R-2SE. The DEFT is considered undefined when the standard error for a simple random sample is zero (when the estimate is close to 0 or 1).

The confidence interval (for example, the one calculated for the variable “households with at least one ITN”) can be interpreted the following way: the overall proportion from the national sample is 0.345 and the standard error is 0.014. To obtain the 95 percent confidence limits, one adds and subtracts twice the standard error to the sample estimate, i.e., 0.345 ± 2 × 0.014. There is a high probability (95 percent) that the true average proportion of households with at least one ITN lies between 0.318 and 0.373.

For the total sample, the value of the design effect (DEFT), averaged over all variables, is 1.95. This means that, due to multi-stage clustering of the sample, the average standard error is increased by a factor of 1.95 over the value observed for a corresponding simple random sample. The actual precision differs from the precision expected during the design of the sample due to several factors: the final size of the sample versus the sample selected; the actual size of DEFT versus the expected; and the actual value of the estimate versus the expected estimate. In addition, the actual precision is different from the expected precision, separately, for each indicator.

The sampling errors are fully described in Appendix B of " Angola Malaria Indicator Survey 2011 - Final Report" pp.49-55.