ABSTRACT
Simple random sampling can be achieved even when no frame is available but where the population members present themselves one at a time as potential sample members. If the sampling frame was constructed in such a way that the population members are listed in random order, then systematic sampling is equivalent to simple random sampling. Once a decision to adopt a stratified sampling scheme is made it becomes necessary to decide how the n units in the sample should be allocated to strata. A special case of this two-stage approach requires that a complete census be taken in each of the primaries selected; this is called cluster sampling. If statisticians are to have a true two-stage sample, that is a sampling scheme for the secondaries within the sampled primaries, they have to face the fact that the primary sampling units can contain very different numbers of secondary sampling units.
