Once an epsem RDD sample is generated by M&RR from the SURVEY sampling system, it is divided into several groups of up to 4,000 records each, with each group, or replicate, having the same geographic proportions.  When calling, the replicates are released and tracked individually.  Each replicate will be called over several days until at least 80% of the sample in the replicate reaches a final status (e.g. has been attempted at least six times, is a complete, a refusal, a disconnected or non-working number).  Once this point is reached, the replicate is considered used, and another replicate is released in its place.  This methodology ensures that, not only is each geographic region given equal representation, but also that the sample is not biased in favor of respondents that are more likely to answer the telephone on a given day.