How Poor Is Your Sample? A Simple Approach for Estimating the Relative Economic Status of Small and Nonrepresentative Samples.
How Poor Is Your Sample? A Simple Approach for Estimating the Relative Economic Status of Small and Nonrepresentative Samples.
复制标题
DOI：
10.9745/ghsp-d-22-00394
复制
发表时间：
2023-04-28
期刊：
影响因子：
--
通讯作者：
Thielman N
中科院分区：
文献类型：
--
作者：
Ostermann J;Hair N;Grzimek V;Zheng S;Gong W;Whetten K;Thielman N
The authors demonstrate the simplicity and utility of a method for estimating the relative economic status of small and nonrepresentative samples relative to existing representative reference populations. Survey data from small and nonrepresentative samples often do not allow for robust inferences about respondents’ relative economic well-being. We demonstrate that out-of-sample predictions based on a straightforward analysis of publicly available data and a small number of purposefully selected survey questions provide robust estimates of relative household wealth for small and nonrepresentative samples. The approach minimizes uncertainty over how to weigh diverse household characteristics and concerns about the precision and external comparability of estimates derived from small and nonrepresentative samples. Program managers should consider this method for describing inclusion and exclusion errors for social assistance programs. Program evaluators could use the method to describe sampling and selection biases in small target populations. Asset-based indices of living standards, or wealth indices, are widely used proxies for economic status; however, such indices are not readily available for small and nonrepresentative samples. We describe a simple out-of-sample prediction approach that uses estimates from large and representative “reference” samples to calculate measures of relative economic status (e.g., wealth index scores) for small and/or nonrepresentative “target” samples. The method relies on the availability of common variables and assumptions about comparable associations between these variables and the underlying construct of interest (e.g., household wealth). We provide 2 sample applications that use Demographic and Health Surveys (DHS) from 5 countries as reference samples. Using ordinary least squares regression, we estimate associations between household characteristics and the DHS wealth index. We use parameter estimates to predict wealth index scores for small nonrepresentative target samples. Comparisons of wealth distributions in the reference and target samples highlight selection effects. Applications of the approach to diverse populations, including populations at high risk of HIV infection and households with orphaned and separated children, demonstrate its usefulness for characterizing the economic status of small and nonrepresentative samples relative to existing reference samples. Women and men in northern Tanzania at high risk of HIV infection were concentrated in the upper half of the wealth distribution. By contrast, the relative distribution of household wealth among households with orphaned and separated children varied greatly across countries and rural versus urban settings. Public health professionals who implement, manage, and evaluate programs in low- and middle-income countries may find this approach applicable because of the simplicity of the estimation methods, low marginal cost of primary data acquisition, and availability of established measures of relative economic status in many publicly available household surveys (e.g., those administered by the DHS Program, World Bank, International Labour Organization, and UNICEF).
影响因子：
2.9
作者：
Ostermann J;Njau B;Hobbie A;Mtuy T;Masaki ML;Shayo A;van Zwetselaar M;Masnick M;Flaherty B;Brown DS;Mühlbacher AC;Thielman NM
通讯作者：
Thielman NM