Abstract:Motivated by the problem of inferring the graph structure of functional connectivity networks from multi-level functional magnetic resonance imaging data, we develop a valid inference framework for high-dimensional graphical models that accounts for group-level heterogeneity. We introduce a neighborhood-based method to learn the graph structure and reframe the problem as that of inferring fixed effect parameters in a doubly high-dimensional linear mixed model. Specifically, we propose a LASSO-based estimator and a de-biased LASSO-based inference framework for the fixed effect parameters in the doubly high-dimensional linear mixed model, leveraging random matrix theory to deal with challenges induced by the identical fixed and random effect design matrices arising in our setting. Moreover, we introduce consistent estimators for the variance components to identify subject-specific edges in the inferred graph. To illustrate the generality of the proposed approach, we also adapt our method to account for serial correlation by learning heterogeneous graphs in the setting of a vector autoregressive model. We demonstrate the performance of the proposed framework using real data and benchmark simulation studies.

What problem does this paper attempt to address?

This paper attempts to address the problem of inferring the graph structure of functional connectivity networks in multi-layer functional magnetic resonance imaging (fMRI) data. Specifically, the authors focus on how to perform effective inference from high-dimensional graphical models while considering inter-group heterogeneity. The main contribution of the paper is the proposal of a method based on a dual high-dimensional linear mixed model, which introduces a neighborhood selection method to learn the graph structure and redefines it as a problem of inferring fixed effect parameters in the dual high-dimensional linear mixed model. ### Main Issues 1. **Inter-group Heterogeneity**: Traditional graphical models usually assume that all subjects have the same dependency structure, but in reality, the functional connectivity network of each subject may differ significantly. 2. **High-dimensional Data**: fMRI data is typically high-dimensional, meaning the number of variables far exceeds the number of samples, making traditional statistical methods difficult to apply directly. 3. **Temporal Correlation**: fMRI data has temporal autocorrelation, requiring special methods to handle this correlation. ### Solutions 1. **Neighborhood Selection Method**: Learn the graph structure through the neighborhood selection method, transforming the problem into inferring fixed effect parameters in the dual high-dimensional linear mixed model. 2. **Dual High-dimensional Linear Mixed Model**: Propose a new dual high-dimensional linear mixed model where the design matrices for fixed and random effects can overlap, addressing the limitations of existing methods in handling overlapping fixed and random effects. 3. **Debiased LASSO Inference Framework**: Use the debiased LASSO method to estimate and infer fixed effect parameters, ensuring consistency and asymptotic normality in high-dimensional settings. 4. **Variance Component Estimation**: Introduce a consistent estimator to identify individual-specific edges, better capturing inter-group heterogeneity. ### Application Scenarios - **Brain Functional Connectivity Networks**: Validate the effectiveness of the proposed method by analyzing resting-state fMRI data from the HCP (Human Connectome Project). - **Other High-dimensional Data**: The method is not only applicable to brain functional connectivity networks but can also be extended to other applications requiring the handling of high-dimensional, multi-layer data, such as the integration of genomic data and data homogenization. ### Main Contributions - **Theoretical Contribution**: Provide a consistent estimation and effective inference framework for the dual high-dimensional linear mixed model, filling the gap in existing methods when dealing with overlapping fixed and random effects. - **Practical Application**: Demonstrate the superior performance of the proposed method in inferring brain functional connectivity networks through real data and simulation experiments. In summary, this paper effectively addresses the problem of inferring the graph structure of functional connectivity networks in multi-layer fMRI data by proposing a new dual high-dimensional linear mixed model and its inference framework, particularly considering inter-group heterogeneity and the characteristics of high-dimensional data.

Inference for Heterogeneous Graphical Models using Doubly High-Dimensional Linear-Mixed Models

High-dimensional Functional Graphical Model Structure Learning via Neighborhood Selection Approach

High-dimensional Mixed Graphical Models

Statistical Inference on High Dimensional Gaussian Graphical Regression Models

Simultaneous Cluster Structure Learning and Estimation of Heterogeneous Graphs for Matrix‐variate Fmri Data

Inference in semiparametric formation models for directed networks

Flexible inference in heterogeneous and attributed multilayer networks

High-Dimensional Undirected Graphical Models for Arbitrary Mixed Data

High Dimensional Semiparametric Latent Graphical Model for Mixed Data

High-Dimensional Joint Estimation of Multiple Directed Gaussian Graphical Models

Estimating Differential Latent Variable Graphical Models with Applications to Brain Connectivity

A Nonparametric Graphical Model for Functional Data With Application to Brain Networks Based on fMRI

Inference for High-Dimensional Linear Mixed-Effects Models: A Quasi-Likelihood Approach

Fair Community Detection and Structure Learning in Heterogeneous Graphical Models

Flexible Regularized Estimation in High-Dimensional Mixed Membership Models

Variable Selection for High-dimensional Nodal Attributes in Social Networks with Degree Heterogeneity

Inference on Large-scale Partially Functional Linear Model with Heterogeneous Errors

Simultaneous Identification of Sparse Structures and Communities in Heterogeneous Graphical Models

Latent Space Model for Higher-order Networks and Generalized Tensor Decomposition

DOUBLY DEBIASED LASSO: HIGH-DIMENSIONAL INFERENCE UNDER HIDDEN CONFOUNDING.

Selective Inference and Learning Mixed Graphical Models