Estimation of sampling errors for survey data

The present chapter provides a brief overview of the various methods used for estimating sampling errors for household survey data generated by various sample designs, ranging from standard designs that can be found in any introductory textbook on sampling theory (for example, Cochran (1977)), to more complex designs used for large-scale household surveys. For the standard sample designs, formulae are provided along with numerical examples to illustrate the estimation of sampling errors, the construction of confidence intervals, and the calculation of design effects and effective sample sizes. Sampling error estimation methods for more complex designs are then presented. The merits and demerits of each method are discussed and numerical examples are provided to illustrate the implementation of the procedures. An example is provided, based on data from a real survey, to illustrate the fact that standard statistical software packages underestimate the sampling errors of survey estimates, leading to wrong conclusions about the parameters of interest to the survey. To avoid this problem, the chapter strongly recommends the use of special statistical software packages that take full account of the complex nature of the designs commonly used for household surveys. Several of these software packages are described and compared.

Related Subject(s): Economic and Social Development