| Sumario: | A new bootstrap proposal, labeled smooth conditional moment (SCM) bootstrap, is introduced for independent but not necessarily identically distributed data, where the classical bootstrap procedure fails. The procedure is shown to encompass the iid and wild bootstrap procedures as special cases. The SCM bootstrap is applied to the construction of confidence intervals for nonparametric equivalence scales estimates, and of critical values for consistent nonparametric test statistics. In both cases, the bootstrap approximations to the small sample properties are usually far superior to those provided by the first-order asymptotic approximations. Reprinted by permission of the publisher.
|
|---|