Abstract:We propose a two-sample testing procedure for high-dimensional time series. To obtain the asymptotic distribution of our l∞-type test statistic under the null hypothesis, we establish high-dimensional central limit theorems (HCLTs) for an α-mixing sequence. Specifically, we derive two HCLTs for the maximum of a sum of high-dimensional α-mixing random vectors under the assumptions of bounded finite moments and exponential tails, respectively. The proposed HCLT for α-mixing sequence under bounded finite moments assumption is novel, and in comparison with existing results, we improve the convergence rate of the HCLT under the exponential tails assumption. To compute the critical value, we employ the blockwise bootstrap method. Importantly, our approach does not require the independence of the two samples, making it applicable for detecting change points in high-dimensional time series. Numerical results emphasize the effectiveness and advantages of our method.
What problem does this paper attempt to address?
### Problems the paper attempts to solve
The paper aims to solve the problem of two - sample tests in high - dimensional time series. Specifically, the author proposes a two - sample test method based on blockwise bootstrap for detecting whether the mean vectors of two high - dimensional time series are equal. This method is especially suitable for dealing with the situation where there is a dependency relationship between the two samples, so that it can effectively detect the change points in high - dimensional time series.
### Background and motivation
In multivariate analysis, a fundamental test problem is to evaluate whether two mean vectors \( \mu_X \) and \( \mu_Y \) are equal. The traditional Hotelling \( T^2 \) test performs well in low - dimensional cases, but in high - dimensional cases, due to the increase in dimension, the performance of traditional methods will decline significantly. In addition, when there is a weak - dependency relationship between two time series, existing methods may not work well.
### Main contributions
1. **High - dimensional central limit theorems**: The author establishes high - dimensional central limit theorems (HCLTs), especially for the maximum of \( \alpha \)-mixing sequences and the maximum of high - dimensional random vector sums. These theorems are derived respectively under the assumptions of finite moments and exponential tails.
2. **Blockwise bootstrap method**: To calculate the critical values, the author adopts the blockwise bootstrap method. This method does not need to assume the independence between the two samples, so it is suitable for detecting change points in high - dimensional time series.
3. **Theoretical results**: The author proves that under the null hypothesis, the distribution of the test statistic \( T_n \) can be effectively approximated by its Gaussian approximation \( T_G^n \). This provides theoretical support for practical applications.
4. **Numerical results**: Through numerical experiments, the author shows the effectiveness and advantages of the proposed method.
### Specific problems
- **Two - sample mean vector test**: Test whether the mean vectors of two high - dimensional time series are equal, that is, \( H_0: \mu_X=\mu_Y \) against \( H_1: \mu_X\neq\mu_Y \).
- **Change - point detection in high - dimensional time series**: Detect the change points in high - dimensional time series, that is, whether the mean vectors change before and after a certain time point \( \tau_0 \).
### Method overview
1. **Test statistic**: Define an \( \ell_\infty \)-type test statistic \( T_n \), in the form of \( T_n=(n_1 n_2)^{1/2}(n_1 + n_2)^{-1/2}|\hat{\mu}_X-\hat{\mu}_Y|_\infty \).
2. **Gaussian approximation**: Approximate the distribution of \( T_n \) by the Gaussian approximation \( T_G^n \), where \( T_G^n = |G|_\infty \) and \( G\sim N(0_p,\Xi_{\tilde{n}}) \).
3. **Blockwise bootstrap method**: Use the blockwise bootstrap method to determine the critical value \( \hat{cv}_\alpha \) at a given significance level \( \alpha \).
### Conclusion
The paper proposes a new two - sample test method based on blockwise bootstrap, which is suitable for mean vector tests and change - point detections in high - dimensional time series. Through theoretical analysis and numerical experiments, the effectiveness and superiority of this method are verified.