Mathematics · Statistics and Probability
Second order asymptotics for M-estimators under non-standard conditions
Nicolas Hengartner, Marten Wegkamp
See on the mapAbstract
This paper establishes, under non-standard conditions, an explicit stochastic approximation of studentized M-estimators 0 n , implicitly defined as solutions to jLi {Xj>) = oi 71 ' 1 ), by a U-statistic U n that is probably concentrated about n in the sense that P[|0 n -U n \ > (nlogn)" 1 ] = o{n~f 2 ). The expansion and concentration hold under weaker smoothness conditions on than those assumed by This approximation is key in rigorously establishing a second order expansion for the sampling distribution of the studentized estimator. Under stronger smoothness assumptions on a similar expansion relates the bootstrap approximation to the true distribution of the studentized M-estimator.