In sample surveys, the analyst is usually interested not only in measuring the mean of a variable for the sample in question, but in estimating the value of this mean for the total population from which the sample was drawn, and the precision with which this estimate of the mean approximates the true mean in the population. If the sample in question is a simple random sample from the population, then the population mean may be taken to be the same as the sample mean, and the variance of this estimate of the mean may be estimated from a standard set of statistical equations. However, in more complex sampling designs involving stratified sampling, multi-stage sampling and cluster sampling, it sometimes becomes very difficult to calculate this variance analytically. In these cases, it is often necessary to resort to other methods of calculating the variance of the parameter estimates. This paper outlines some methods of variance estimation, and concentrates on the use of design effects and, particularly, replication as a means of variance estimation. These methods are illustrated by reference to a hypothetical survey of community attitudes towards the upgrading of a county airport.
Abstract