Abstract:Let $\bm{x}_1,\cdots,\bm{x}_n$ be a random sample of size $n$ from a $p$-dimensional population distribution, where $p=p(n)\rightarrow\infty$. Consider a symmetric matrix $W=X^\top X$ with parameters $n$ and $p$, where $X=(\bm{x}_1,\cdots,\bm{x}_n)^\top$. In this paper, motivated by model selection theory in high-dimensional statistics, we mainly investigate the asymptotic behavior of the eigenvalues of the principal minors of the random matrix $W$. For the Gaussian case, under a simple condition that $m=o(n/\log p)$, we obtain the asymptotic results on maxima and minima of the eigenvalues of all $m\times m$ principal minors of $W$. We also extend our results to general distributions with some moment conditions. Moreover, we gain the asymptotic results of the extreme eigenvalues of the principal minors in the case of the real Wigner matrix. Finally, similar results for the maxima and minima of the eigenvalues of all the principal minors with a size smaller than or equal to $m$ are also given.

What problem does this paper attempt to address?

### Problems Addressed by the Paper This paper primarily explores the asymptotic behavior of the extreme eigenvalues of principal submatrices of random matrices in the context of model selection theory in high-dimensional statistics. Specifically, the authors investigate the asymptotic properties of the extreme eigenvalues (maximum and minimum) of all $ m \times m $ principal submatrices of a random matrix $ W = X^\top X $ under certain moment conditions. #### Main Research Content: 1. **Results in the Gaussian Case**: For Gaussian distributions, when $ m = o(n / \log p) $, the authors obtained asymptotic results for the maximum and minimum eigenvalues of all $ m \times m $ principal submatrices. 2. **General Distribution Case**: The authors extended the results to general distributions with certain moment conditions, further exploring the asymptotic properties of the extreme eigenvalues under these distributions. 3. **Real Wigner Matrix Case**: Additionally, the authors studied the asymptotic results of the extreme eigenvalues of real Wigner matrices. #### Research Motivation: - **Variable Selection in High-Dimensional Statistics**: In high-dimensional statistics, variable selection is an important problem. The sparse Riesz condition (SRC) is one of the key assumptions for variable selection methods (such as the MC+ method), and SRC is closely related to the extreme eigenvalues of principal submatrices of random matrices. - **Consistency of Model Selection**: By studying the asymptotic behavior of extreme eigenvalues, the consistency and unbiasedness of model selection methods can be verified. #### Research Significance: - **Theoretical Contribution**: This paper relaxes some conditions in the existing literature, such as $ m = o(\min\{(\log p)^{1/3} / \log \log p, n^{1/4} / (\log n)^{3/2} (\log p)^{1/2}\}) $, and obtains asymptotic results under broader distribution conditions. - **Application Prospects**: These results have important implications for model selection, principal component analysis, factor analysis, and other fields in high-dimensional data. In summary, this paper provides new theoretical foundations and methods for model selection theory in high-dimensional statistics by deeply studying the asymptotic behavior of the extreme eigenvalues of principal submatrices of random matrices.

Extreme eigenvalues of principal minors of random matrix with moment conditions

Largest Eigenvalues of Principal Minors of Deformed Gaussian Orthogonal Ensembles and Wishart Matrices

Extreme eigenvalues of Laplacian random matrices with Gaussian entries

On asymptotics of eigenvectors of large sample covariance matrix

Large deviations of extreme eigenvalues of generalized sample covariance matrices

Extreme eigenvalues of sample covariance matrices under generalized elliptical models with applications

Almost sure convergence of the largest and smallest eigenvalues of high-dimensional sample correlation matrices

The Asymptotic Properties of the Extreme Eigenvectors of High-dimensional Generalized Spiked Covariance Model

Wegner estimate and upper bound on the eigenvalue condition number of non-Hermitian random matrices

Moderate Deviations for Extreme Eigenvalues of Real-Valued Sample Covariance Matrices

Concentration and moment inequalities for heavy-tailed random matrices

Asymptotic Normality of the Largest Eigenvalue for Noncentral Sample Covariance Matrices

Extreme Eigenvalues of Large-Dimensional Spiked Fisher Matrices with Application

Extreme value analysis for the sample autocovariance matrices of heavy-tailed multivariate time series

Moment Approach for Singular Values Distribution of a Large Auto-Covariance Matrix

Principal minors of Gaussian orthogonal ensemble

Extremal eigenvalues of sample covariance matrices with general population

Exact Separation of Eigenvalues of Large Dimensional Noncentral Sample Covariance Matrices

Extremal random matrices with independent entries and matrix superconcentration inequalities

Extremal Eigenvalues of Random Kernel Matrices with Polynomial Scaling

Outliers for deformed inhomogeneous random matrices