1: \begin{abstract}

2: Zou [\textit{J. Amer. Statist. Assoc.} \textbf{101} (2006) 1418--1429]

3: proposed the Adaptive LASSO

4: (ALASSO) method for simultaneous variable selection and

5: estimation of the regression parameters, and

6: established its oracle property. In this paper, we

7: investigate the rate of convergence of the

8: ALASSO estimator to the oracle distribution

9: when the dimension of the regression parameters

10: %is undounded.

11: may grow to infinity with the sample size.

12: It is shown that the rate critically depends on the choices of

13: the penalty parameter and the initial estimator,

14: among other factors, and that confidence intervals

15: (CIs) based on the oracle limit law often have poor

16: coverage accuracy. As an alternative, we consider

17: % higher order properties of

18: the residual bootstrap method for the ALASSO

19: estimators %of the nonzero regression parameters

20: that has been recently shown to be consistent; cf.

21: Chatterjee and Lahiri [\textit{J. Amer. Statist. Assoc.} \textbf{106}

22: (2011a) 608--625].

23: %;\textit{J. Amer. Statist. Assoc.}).

24: We show that the bootstrap applied

25: to a suitable studentized version of the

26: ALASSO estimator achieves second-order correctness,

27: even when the dimension of the regression parameters

28: is unbounded.

29: %Further, using higher order expansions, we

30: %develop a new bias-adjusted studentized version of

31: %the ALASSO estimator and show that the bootstrap

32: %approximation to the distribution of this modified

33: %pivot attains the same optimal level of accuracy as

34: %in the traditional case where the dimensions

35: % of the parameters id fixed

36: Results from a moderately large simulation study show

37: marked improvement in coverage accuracy for the bootstrap

38: CIs over the oracle based CIs.

39: \end{abstract}