independent
Independent pilot sampling for groups in multilevel BLUE
Specification
Alias: None
Arguments: None
Description
The pilot sample provides initial estimates of group covariances during the first iteration of the multilevel best linear unbiased estimator (ML BLUE) sampling approach.
ML BLUE requires independent samples per group, and this option starts from independent pilot samples for each of the groups in ML BLUE. This approach results in a more expensive initialization for the group covariances than the shared pilot sample approach (the default) and it may result in lower efficiency in the case where it invests pilot samples for groups that turn out to be inactive in the subsequent resource allocations. Its advantage is simplicity in that group sample increments simply update the existing group covariances since they were constructed independently from the beginning.
