1: \begin{abstract}

2:   Let $G$ be a non--linear function of a Gaussian process

3:   $\{X_t\}_{t\in\mathbb{Z}}$ with long--range dependence. The resulting process

4:   $\{G(X_t)\}_{t\in\mathbb{Z}}$ is not Gaussian when $G$ is not linear. We

5:   consider random wavelet coefficients associated with

6:   $\{G(X_t)\}_{t\in\mathbb{Z}}$ and the corresponding wavelet scalogram which

7:   is the average of squares of wavelet coefficients over locations. We obtain

8:   the asymptotic behavior of the scalogram as the number of observations and

9:   the analyzing scale tend to infinity. It is known that when $G$ is a Hermite

10:   polynomial of any order, then the limit is either the Gaussian or the

11:   Rosenblatt distribution, that is, the limit can be represented by a multiple

12:   Wiener-Itô integral of order one or two. We show, however, that there are

13:   large classes of functions $G$ which yield a higher order Hermite

14:   distribution, that is, the limit can be represented by a a multiple

15:   Wiener-Itô integral of order greater than two. This happens for example if

16:   $G$ is a linear combination of a Hermite polynomial of order $1$ and a

17:   Hermite polynomial of order $q>3$. The limit in this case can be Gaussian but

18:   it can also be a Hermite distribution of order $q-1>2$. This depends not only

19:   on the relation between the number of observations and the scale size but

20:   also on whether $q$ is larger or smaller than a new critical index $q^*$. The

21:   convergence of the wavelet scalogram is therefore significantly more complex

22:   than the usual one.

23:  \end{abstract}