Search

While widely used as a general method for uncertainty quantification, the bootstrap method encounters difficulties that raise concerns about its validity in practical applications. This paper introduces a new resampling-based method, termed $\textit{calibrated bootstrap}$, designed to generate finite sample-valid parametric inference from a sample of size $n$. The central idea is to calibrate an $m$-out-of-$n$ resampling scheme, where the calibration parameter $m$ is determined against inferential pivotal quantities derived from the cumulative distribution functions of loss functions in parameter estimation. The method comprises two algorithms. The first, named $\textit{resampling approximation}$ (RA), employs a $\textit{stochastic approximation}$ algorithm to find the value of the calibration parameter $m=m_\alpha$ for a given $\alpha$ in a manner that ensures the resulting $m$-out-of-$n$ bootstrapped $1-\alpha$ confidence set is valid. The second algorithm, termed $\textit{distributional resampling}$ (DR), is developed to further select samples of bootstrapped estimates from the RA step when constructing $1-\alpha$ confidence sets for a range of $\alpha$ values is of interest. The proposed method is illustrated and compared to existing methods using linear regression with and without $L_1$ penalty, within the context of a high-dimensional setting and a real-world data application. The paper concludes with remarks on a few open problems worthy of consideration.

Obesity is widely recognized as a critical and pervasive health concern. We strive to identify important genetic risk factors from hundreds of thousands of single nucleotide polymorphisms (SNPs) for obesity. We propose and apply a novel Quantile Regression with Insight Fusion (QRIF) approach that can integrate insights from established studies or domain knowledge to simultaneously select variables and modeling for ultra-high dimensional genetic data, focusing on high conditional quantiles of body mass index (BMI) that are of most interest. We discover interesting new SNPs and shed new light on a comprehensive view of the underlying genetic risk factors for different levels of BMI. This may potentially pave the way for more precise and targeted treatment strategies. The QRIF approach intends to balance the trade-off between the prior insights and the observed data while being robust to potential false information. We further establish the desirable asymptotic properties under the challenging non-differentiable check loss functions via Huber loss approximation and nonconvex SCAD penalty via local linear approximation. Finally, we develop an efficient algorithm for the QRIF approach. Our simulation studies further demonstrate its effectiveness., Comment: This article is submitted to Journal of the American Statistical Association

The patterns of particulate matter with diameters that are generally 2.5 micrometers and smaller (PM2.5) are heterogeneous in China nationwide but can be homogeneous region-wide. To reduce the adverse effects from PM2.5, policymakers need to develop location-specific regulations based on nationwide clustering analysis of PM2.5 concentrations. However, such an analysis is challenging because the data have complex structures and are usually noisy. In this study, we propose a robust clustering framework using a novel concept of depth, which can handle both magnitude and shape outliers effectively and incorporate spatial information. We apply this framework to a PM2.5 dataset and reveal ten regions in China that have distinct PM2.5 patterns, and the homogeneity within each cluster is also confirmed. The clusters have clearly visible boundaries and enable policymakers to develop local emission control policies and establish regional collaborative systems to control air pollution in China.

Reconstruction of interaction network between random events is a critical problem arising from statistical physics and politics to sociology, biology, and psychology, and beyond. The Ising model lays the foundation for this reconstruction process, but finding the underlying Ising model from the least amount of observed samples in a computationally efficient manner has been historically challenging for half a century. By using the idea of sparsity learning, we present a approach named SIMPLE that has a dominant sample complexity from theoretical limit. Furthermore, a tuning-free algorithm is developed to give a statistically consistent solution of SIMPLE in polynomial time with high probability. On extensive benchmarked cases, the SIMPLE approach provably reconstructs underlying Ising models with global optimality. The application on the U.S. senators voting in the last six congresses reveals that both the Republicans and Democrats noticeably assemble in each congresses; interestingly, the assembling of Democrats is particularly pronounced in the latest congress.

Analysis of high-dimensional data has led to increased interest in both single index models (SIMs) and best subset selection. SIMs provide an interpretable and flexible modeling framework for high-dimensional data, while best subset selection aims to find a sparse model from a large set of predictors. However, best subset selection in high-dimensional models is known to be computationally intractable. Existing methods tend to relax the selection, but do not yield the best subset solution. In this paper, we directly tackle the intractability by proposing the first provably scalable algorithm for best subset selection in high-dimensional SIMs. Our algorithmic solution enjoys the subset selection consistency and has the oracle property with a high probability. The algorithm comprises a generalized information criterion to determine the support size of the regression coefficients, eliminating the model selection tuning. Moreover, our method does not assume an error distribution or a specific link function and hence is flexible to apply. Extensive simulation results demonstrate that our method is not only computationally efficient but also able to exactly recover the best subset in various settings (e.g., linear regression, Poisson regression, heteroscedastic models).

In a region $R$ consisting of unit squares, a (domino) tiling is a collection of dominoes (the union of two adjacent squares) which pave fully the region. The flip graph of $R$ is defined on the set of all tilings of $R$ where two tilings are adjacent if we change one from the other by a flip (a $90^{\circ}$ rotation of a pair of side-by-side dominoes). Let $n\geq 1$ and $m\geq 2$ be integers. In a recent paper it was proved that the flip graph of $(2n+1)\times 2m$ quadriculated torus consists of two isomorphic components. In this paper, we generalize the result to the toroidal grid $T(2n+1,2m,2r)$ which is obtained from an $(2n+1)\times 2m$ chessboard by sticking the left and right sides and then identifying the top and bottom sides with a torsion $2r$ squares for $1\leq r\leq m$. For a tiling $t$, we associate an integer called forcing number as the minimum number of dominoes in $t$ that is contained in no other tiling. As an application, we obtain that the forcing numbers of all tilings of $T(2n+1,2m,2r)$ form an integer-interval. Moreover, we prove that the maximum value of the interval is $\frac{m(2n+1)}{2}$ if $\frac{m}{(r,m)}$ is even, and $\frac{m(2n+1)+(r,m)}{2}$ otherwise.

A graph $G$ with four or more vertices is called bicritical if the removal of any pair of distinct vertices of $G$ results in a graph with a perfect matching. A bicritical graph is minimal if the deletion of each edge results in a non-bicritical graph. Recently, Y. Zhang et al. and F. Lin et al. respectively showed that bicritical graphs without removable edges and minimal bricks have at least four cubic vertices. In this note, we show that minimal bicritical graphs also have at least four cubic vertices, so confirming O. Favaron and M. Shi's conjecture in the case of $k=2$ on minimal $k$-factor critical graphs., Comment: 9 pages, 2 figures

We obtain the formulae for the numbers of 4-matchings and 5-matchings in terms of the number of hexagonal faces in (4, 6)-fullerene graphs by studying structural classification of 6-cycles and some local structural properties, which correct the corresponding wrong results published. Furthermore, we obtain a formula for the number of 6-matchings in tubular (4, 6)-fullerenes in terms of the number of hexagonal faces, and a formula for the number of 6-matchings in the other (4,6)-fullerenes in terms of the numbers of hexagonal faces and dual-squares., Comment: This article was already published in 2017 in MATCH Commun. Math. Comput. Chem. We are uploading it to arXiv for readers' convenience

Last decade, numerous giant data center networks are built to provide increasingly fashionable web applications. For two integers $m\geq 0$ and $n\geq 2$, the $m$-dimensional DCell network with $n$-port switches $D_{m,n}$ and $n$-dimensional BCDC network $B_{n}$ have been proposed. Connectivity is a basic parameter to measure fault-tolerance of networks. As generalizations of connectivity, structure (substructure) connectivity was recently proposed. Let $G$ and $H$ be two connected graphs. Let $\mathcal{F}$ be a set whose elements are subgraphs of $G$, and every member of $\mathcal{F}$ is isomorphic to $H$ (resp. a connected subgraph of $H$). Then $H$-structure connectivity $\kappa(G; H)$ (resp. $H$-substructure connectivity $\kappa^{s}(G; H)$) of $G$ is the size of a smallest set of $\mathcal{F}$ such that the rest of $G$ is disconnected or the singleton when removing $\mathcal{F}$. Then it is meaningful to calculate the structure connectivity of data center networks on some common structures, such as star $K_{1,t}$, path $P_k$, cycle $C_k$, complete graph $K_s$ and so on. In this paper, we obtain that $\kappa (D_{m,n}; K_{1,t})=\kappa^s (D_{m,n}; K_{1,t})=\lceil \frac{n-1}{1+t}\rceil+m$ for $1\leq t\leq m+n-2$ and $\kappa (D_{m,n}; K_s)= \lceil\frac{n-1}{s}\rceil+m$ for $3\leq s\leq n-1$ by analyzing the structural properties of $D_{m,n}$. We also compute $\kappa(B_n; H)$ and $\kappa^s(B_n; H)$ for $H\in \{K_{1,t}, P_{k}, C_{k}|1\leq t\leq 2n-3, 6\leq k\leq 2n-1 \}$ and $n\geq 5$ by using $g$-extra connectivity of $B_n$.

Sparse reduced rank regression is an essential statistical learning method. In the contemporary literature, estimation is typically formulated as a nonconvex optimization that often yields to a local optimum in numerical computation. Yet, their theoretical analysis is always centered on the global optimum, resulting in a discrepancy between the statistical guarantee and the numerical computation. In this research, we offer a new algorithm to address the problem and establish an almost optimal rate for the algorithmic solution. We also demonstrate that the algorithm achieves the estimation with a polynomial number of iterations. In addition, we present a generalized information criterion to simultaneously ensure the consistency of support set recovery and rank estimation. Under the proposed criterion, we show that our algorithm can achieve the oracle reduced rank estimation with a significant probability. The numerical studies and an application in the ovarian cancer genetic data demonstrate the effectiveness and scalability of our approach., Comment: 38 pages, 5 figures

In a region $R$ consisting of unit squares, a domino is the union of two adjacent squares and a (domino) tiling is a collection of dominoes with disjoint interior whose union is the region. The flip graph $\mathcal{T}(R)$ is defined on the set of all tilings of $R$ such that two tilings are adjacent if we change one to another by a flip (a $90^{\circ}$ rotation of a pair of side-by-side dominoes). It is well-known that $\mathcal{T}(R)$ is connected when $R$ is simply connected. By using graph theoretical approach, we show that the flip graph of $2m\times(2n+1)$ quadriculated cylinder is still connected, but the flip graph of $2m\times(2n+1)$ quadriculated torus is disconnected and consists of exactly two isomorphic components. For a tiling $t$, we associate an integer $f(t)$, forcing number, as the minimum number of dominoes in $t$ that is contained in no other tilings. As an application, we obtain that the forcing numbers of all tilings in $2m\times (2n+1)$ quadriculated cylinder and torus form respectively an integer interval whose maximum value is $(n+1)m$.

A graph $G$ of order $n$ is said to be $k$-factor-critical for integers $1\leq k < n$, if the removal of any $k$ vertices results in a graph with a perfect matching. A $k$-factor-critical graph $G$ is called minimal if for any edge $e\in E(G)$, $G-e$ is not $k$-factor-critical. In 1998, O. Favaron and M. Shi conjectured that every minimal $k$-factor-critical graph of order $n$ has the minimum degree $k+1$ and confirmed it for $k=1, n-2, n-4$ and $n-6$. By using a novel approach, we have confirmed it for $k = n - 8$ in a previous paper. Continuing this method, we prove the conjecture to be true for $k=n-10$ in this paper., Comment: 21 pages,14 figures

Non-Euclidean data become more prevalent in practice, necessitating the development of a framework for statistical inference analogous to that for Euclidean data. Quantile is one of the most important concepts in traditional statistical inference; we introduce the counterpart, both locally and globally, for data objects in metric spaces. This is realized by expanding upon the metric distribution function proposed by Wang et al. (2021). Rank and sign are defined at local and global levels as a natural consequence of the center-outward ordering of metric spaces brought about by the local and global quantiles. The theoretical properties are established, such as the root-$n$ consistency and uniform consistency of the local and global empirical quantiles and the distribution-freeness of ranks and signs. The empirical metric median, which is defined here as the 0th empirical global metric quantile, is proven to be resistant to contamination by means of both theoretical and numerical approaches. Quantiles have been shown to be valuable through extensive simulations in a number of metric spaces. Moreover, we introduce a family of fast rank-based independence tests for a generic metric space. Monte Carlo experiments show good finite-sample performance of the test.

A graph $G$ of order $n$ is said to be $k$-factor-critical for integers $1\leq k < n$, if the removal of any $k$ vertices results in a graph with a perfect matching. $1$- and $2$-factor-critical graphs are the well-known factor-critical and bicritical graphs, respectively. A $k$-factor-critical graph $G$ is called minimal if for any edge $e\in E(G)$, $G-e$ is not $k$-factor-critical. In 1998, O. Favaron and M. Shi conjectured that every minimally $k$-factor-critical graph of order $n$ has the minimum degree $k+1$ and confirmed it for $k=1, n-2, n-4$ and $n-6$. In this paper, we use a simple method to reprove the above result. As a main result, the further use of this method enables ones to prove the conjecture to be true for $k=n-8$. We also obtain that every minimally $(n-6)$-factor-critical graph of order $n$ has at most $n-\Delta(G)$ vertices with the maximum degree $\Delta(G)$ for $n-4\leq \Delta(G)\leq n-1$., Comment: 20

A graph is $1$-$planar$ if it can be drawn in the plane so that each edge is crossed by at most one other edge. Moreover, a 1-planar graph $G$ is $optimal$ if it satisfies $|E(G)|=4|V(G)|-8$. J. Fujisawa et al. [16] first considered matching extension of optimal 1-planar graphs, obtained that each optimal 1-planar graph of even order is 1-extendable and characterized 2-extendable optimal 1-planar graphs and 3-matchings extendable to perfect matchings as well. In this short paper, we prove that no optimal $1$-planar graph is 3-extendable. Further we mainly obtain that no 1-planar graph is 5-extendable by the discharge method and also construct a 4-extendable 1-planar graph. Finally we get that no 1-planar graph is 7-factor-critical and no optimal 1-planar graph is 6-factor-critical., Comment: 13 pages, 8 figures

The connectivity is an important parameter to evaluate the robustness of a network. As a generalization, structure connectivity and substructure connectivity of graphs were proposed. For connected graphs $G$ and $H$, the $H$-structure connectivity $\kappa(G; H)$ (resp. $H$-substructure connectivity $\kappa^{s}(G; H)$) of $G$ is the minimum cardinality of a set of subgraphs $F$ of $G$ that each is isomorphic to $H$ (resp. to a connected subgraph of $H$) so that $G-F$ is disconnected or the singleton. As popular variants of hypercubes, the $n$-dimensional folded hypercubes $FQ_{n}$ and augmented cubes $AQ_{n}$ are attractive interconnected network prototypes for multiple processor systems. In this paper, we obtain that $\kappa(FQ_{n};K_{1,m})=\kappa^{s}(FQ_{n};K_{1,m})=\lceil\frac{n+1}{2}\rceil$ for $2\leqslant m\leqslant n-1$, $n\geqslant 7$, and $\kappa(AQ_{n};K_{1,m})=\kappa^{s}(AQ_{n};K_{1,m})=\lceil\frac{n-1}{2}\rceil$ for $4\leqslant m\leqslant \frac{3n-15}{4}$.

Distribution function is essential in statistical inference, and connected with samples to form a directed closed loop by the correspondence theorem in measure theory and the Glivenko-Cantelli and Donsker properties. This connection creates a paradigm for statistical inference. However, existing distribution functions are defined in Euclidean spaces and no longer convenient to use in rapidly evolving data objects of complex nature. It is imperative to develop the concept of distribution function in a more general space to meet emerging needs. Note that the linearity allows us to use hypercubes to define the distribution function in a Euclidean space, but without the linearity in a metric space, we must work with the metric to investigate the probability measure. We introduce a class of metric distribution functions through the metric between random objects and a fixed location in metric spaces. We overcome this challenging step by proving the correspondence theorem and the Glivenko-Cantelli theorem for metric distribution functions in metric spaces that lie the foundation for conducting rational statistical inference for metric space-valued data. Then, we develop homogeneity test and mutual independence test for non-Euclidean random objects, and present comprehensive empirical evidence to support the performance of our proposed methods.