Abstract:We study the effect of observing a stationary process at irregular time points via a renewal process. We establish a sharp difference in the asymptotic behaviour of the self-normalized sample mean of the observed process depending on the renewal process. In particular, we show that if the renewal process has a moderate heavy tail distribution then the limit is a so-called Normal Variance Mixture (NVM) and we characterize the randomized variance part of the limiting NVM as an integral function of a Lévy stable motion. Otherwise, the normalized sample mean will be asymptotically normal.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the asymptotic behavior of the self - normalized sample mean of the observed process when observing a long - memory stationary process at irregular time points. Specifically, the paper studies the influence of observing a long - memory stationary process at irregular time points through a renewal process, and explores the influence of different tail distributions of the renewal process on the asymptotic distribution of the self - normalized sample mean of the observed process.
### Main research questions:
1. **Asymptotic behavior under different renewal processes**: The paper focuses on the asymptotic behavior of the self - normalized sample mean of the observed process \( Y_k \) when using renewal processes with different tail distributions to model non - uniformly observed data.
2. **Differences in limit distributions**: In particular, the paper shows that if the renewal process has a moderately heavy - tailed distribution, the limit distribution is a so - called normal variance - mixture (NVM), and further characterizes the random variance part in the limit NVM as an integral function of a Lévy stable motion. Otherwise, the normalized sample mean will be asymptotically normal.
### Specific problem descriptions:
- **Light - tailed or moderately heavy - tailed renewal processes**: If the renewal process has finite moments (such as the Poisson process) or a very heavy - tailed distribution, the limit distribution is a normal distribution.
- **Moderately heavy - tailed renewal processes**: If the renewal process has an intermediate heavy - tailed distribution, the limit distribution becomes a normal variance - mixture (NVM), which is the product of two independent random variables, one of which is positive (usually called random variance) and the other is a standard normal distribution.
### Main contributions of the paper:
- **Exact characterization of asymptotic behavior**: The paper establishes the exact differences in the asymptotic behavior of the self - normalized sample mean of the observed process \( Y_k \) under different renewal processes.
- **Complete characterization of random variance**: The paper provides a complete characterization of the random variance part in the limit NVM as an integral function of a Lévy stable sub - process.
- **Continuity results**: The paper also establishes the continuity of the limit distribution during the transition from very heavy - tailed to moderately heavy - tailed.
### Key formulas:
- **Covariance function of the long - memory process**:
\[
\sigma_X(h)\sim\tilde{C}_d h^{2d - 1},\quad\text{as }h\to\infty,\quad0 < d <\frac{1}{2}
\]
- **Covariance function of the observed process**:
\[
\sigma_Y(h)\sim\begin{cases}
\tilde{C}_d\frac{\Gamma\left(\frac{1 - 2d}{\alpha}\right)}{\alpha\Gamma(1 - 2d)}b_h^{2d - 1},&\text{if }\alpha < 1,\\
\tilde{C}_d b_h^{2d - 1}\left(\frac{\ell^*(b_h)}{\ell(b_h)}\right)^{2d - 1},&\text{if }\alpha = 1.
\end{cases}
\]
- **Random variance in normal variance - mixture (NVM)**:
\[
Z(\alpha,d)=C_{\alpha,1 - 2d}\int_{[0,1]^2}|L_x - L_y|^{2d - 1}\,dx\,dy
\]
where \( L_t \) is a Lévy stable motion with parameter \(\alpha\), and
\[
C_{\alpha,1 - 2d}=\frac{(\alpha + 2d - 1)(2\alpha + 2d - 1)}{2\alpha^2}\frac{\Gamma(1 - 2d)}{\Gamma\left(\frac{1 - 2d}{\alpha}\right)}
\]
Through these studies, the paper provides an important theoretical basis for understanding the statistical properties of long - memory processes observed at irregular time points.