1: \begin{abstract}

2: The estimation of the L\'evy density,

3: the infinite-dimensional parameter controlling the

4: jump dynamics of a L\'evy process, is considered here

5: under a discrete-sampling scheme. In this setting,

6: the jumps are latent variables, the statistical

7: properties of which can be assessed when the frequency and time

8: horizon of

9: observations increase to infinity at suitable rates.

10: Nonparametric estimators for the L\'evy density

11: based on \textit{Grenander's method of sieves}

12: was proposed in Figueroa-L\'opez [\textit{IMS Lecture Notes} \textbf

13: {57} (2009) 117--146].

14: In this paper, central limit theorems for these sieve estimators, both

15: pointwise and uniform on

16: an interval away from the origin,

17: are obtained, leading to pointwise confidence intervals and bands for the

18: L\'evy density. In the pointwise case,

19: our estimators converge to the L\'evy density

20: at a rate that is arbitrarily close to the rate of the minimax risk of

21: estimation on smooth L\'evy densities.

22: In the case of uniform bands and discrete regular sampling, our

23: results are consistent with the case of density estimation, achieving a

24: rate of order arbitrarily close to $\log^{-1/2}(n)

25: \cdot n^{-1/3}$, where $n$ is the number of observations. The

26: convergence rates are valid,

27: provided that $s$ is smooth enough and that the time horizon

28: $T_{n}$ and the dimension of the sieve are appropriately chosen in

29: terms of $n$.

30: 

31: \end{abstract}