Abstract:We derive functional convergence of the partial maxima stochastic processes of multivariate linear processes with weakly dependent heavy-tailed innovations and random coefficients. The convergence takes place in the space of $\mathbb{R}^{d}$--valued càdlàg functions on $[0,1]$ endowed with the weak Skorokhod $M_{1}$ topology. We also show that this topology in general can not be replaced by the standard (or strong) $M_{1}$ topology.
What problem does this paper attempt to address?
The paper attempts to address the problem of functional convergence of the partial maximum process of multivariate linear processes under heavy-tailed innovations and random coefficients. Specifically, the authors study the functional convergence of the partial maximum process of multivariate linear processes with weakly dependent heavy-tailed innovations and random coefficients under the weak Skorokhod M1 topology.
### Background and Motivation
1. **Classical Extreme Value Theory**: For an independent and identically distributed (i.i.d.) sequence \( \{X_i\} \), the asymptotic distribution of the partial maximum \( M_n = \max\{X_1, X_2, \ldots, X_n\} \) is the extreme value distribution.
2. **Weak Dependence Case**: When the sequence \( \{X_i\} \) has weak dependence, the functional convergence of the partial maximum process becomes complex. Previous studies have shown that for certain types of weakly dependent processes, such as linear processes or moving average processes, the partial maximum process converges under the J1 topology.
3. **Multivariate Case**: In the multivariate case, there is less research on the functional convergence of the partial maximum process, especially when the innovations have heavy-tailed distributions and the coefficients are random.
### Research Objectives
1. **Extend Existing Results**: Generalize the functional convergence results of the partial maximum process of univariate linear processes to multivariate linear processes.
2. **Consider Weak Dependence and Random Coefficients**: Study the functional convergence of the partial maximum process of multivariate linear processes with weakly dependent heavy-tailed innovations and random coefficients under the weak Skorokhod M1 topology.
3. **Verify Topology Choice**: Demonstrate that in general cases, the weak Skorokhod M1 topology cannot be replaced by the standard (strong) M1 topology.
### Main Contributions
1. **Functional Convergence**: By using point process convergence and the continuous mapping theorem, the paper proves the functional convergence of the partial maximum process of multivariate linear processes with weakly dependent heavy-tailed innovations and random coefficients under the weak Skorokhod M1 topology.
2. **Importance of Topology Choice**: Through examples, the paper illustrates that in some cases, the weak Skorokhod M1 topology is necessary, while the standard M1 topology is not applicable.
### Methods and Techniques
1. **Point Process Theory**: Utilizing point process convergence theory, the problem of functional convergence of the partial maximum process is transformed into a problem of point process convergence.
2. **Continuous Mapping Theorem**: Applying the continuous mapping theorem, the convergence results of point processes are converted into functional convergence results of the partial maximum process.
3. **Weak Skorokhod M1 Topology**: The weak Skorokhod M1 topology is chosen because, under this topology, the functional convergence of the partial maximum process is easier to handle, especially in the presence of large value clusters.
### Conclusion
The paper successfully proves the functional convergence of the partial maximum process of multivariate linear processes with weakly dependent heavy-tailed innovations and random coefficients under the weak Skorokhod M1 topology. It also illustrates the necessity of the weak Skorokhod M1 topology through examples. These results provide new insights into the extreme value theory of multivariate linear processes and lay the foundation for further research in related fields.