One-step Bootstrapping for Smooth Iterative Procedures

摘要: SUMMARY Resampling techniques have the potential to provide useful information about sampling distribution of estimators many population characteristics. Ambitious schemes such as bootstrap and iterated imply a substantial increase in computational effort. For some iterative procedures, generalized least squares or EM algorithm, it is possible avoid fully iterating each replication convergence. By analysing expansions defining equation, we can extract asymptotically correct estimates from single step for replication. In this paper demonstrate large sample validity computationally efficient approach illustrate its small applicability. Whether not adjustment represents an adequate replacement full iteration depends on nature problem desired accuracy quantiles. If subsequent iterations are adjusted, then greater enhancement rate achieved practical significant.