Abstract:Gaussian graphical regressions have emerged as a powerful approach for regressing the precision matrix of a Gaussian graphical model on covariates, which, unlike traditional Gaussian graphical models, can help determine how graphs are modulated by high dimensional subject-level covariates, and recover both the population-level and subject-level graphs. To fit the model, a multi-task learning approach {achieves} %has been shown to result in lower error rates compared to node-wise regressions. However, due to the high complexity and dimensionality of the Gaussian graphical regression problem, the important task of statistical inference remains unexplored. We propose a class of debiased estimators based on multi-task learners for statistical inference in Gaussian graphical regressions. We show that debiasing can be performed quickly and separately for the multi-task learners. In a key debiasing step {that estimates} %involving the estimation of the inverse covariance matrix, we propose a novel {projection technique} %diagonalization approach that dramatically reduces computational costs {in optimization} to scale only with the sample size $n$. We show that our debiased estimators enjoy a fast convergence rate and asymptotically follow a normal distribution, enabling valid statistical inference such as constructing confidence intervals and performing hypothesis testing. Simulation studies confirm the practical utility of the proposed approach, and we further apply it to analyze gene co-expression graph data from a brain cancer study, revealing meaningful biological relationships.
What problem does this paper attempt to address?
### Problems the paper attempts to solve
This paper aims to solve the statistical inference problems in high - dimensional Gaussian graphical regression models. Specifically, the Gaussian graphical regression model is a powerful method for regressing the precision matrix of the Gaussian graphical model with covariates, which helps to determine how the graph is modulated by high - dimensional individual - level covariates and recover the group - level and individual - level graph structures. However, due to the high complexity and dimensionality of the Gaussian graphical regression problem, the statistical inference task has not been fully explored.
### Main contributions
1. **Debiased estimators**: The authors propose a class of debiased estimators based on multi - task learners for statistical inference in Gaussian graphical regression models. These debiased estimators can be quickly and independently applied to multi - task learners.
2. **Projection techniques**: In the crucial debiasing step, the authors propose a new projection technique, which significantly reduces the computational cost in the optimization process, making it vary only with the sample size \(n\).
3. **Theoretical results**: The authors prove that the debiased estimators have a fast convergence rate and asymptotically follow a normal distribution, enabling effective statistical inferences such as constructing confidence intervals and performing hypothesis tests.
4. **Practical applications**: Through simulation studies and the practical analysis of brain cancer gene co - expression graph data, the effectiveness and practicality of the proposed method are verified, revealing meaningful biological relationships.
### Method overview
1. **Model description**:
- The Gaussian graphical regression model considers a random vector \(X=(X_1,\ldots,X_p)\) that follows a multivariate normal distribution \(N_p(0,\Sigma)\), where \(\Sigma\in\mathbb{R}^{p\times p}\) is the covariance matrix and \(\Omega = \Sigma^{-1}\) is the precision matrix.
- Covariates \(U=(U_1,\ldots,U_q)^{\top}\) are introduced into the model, and the conditional distribution is \(X\mid U = u\sim N_p(\mu(u),\Omega^{-1}(u))\), where \(\mu(u)=\Gamma u\) and \(\Omega(u)=B_0+\sum_{h = 1}^q B_h u_h\).
2. **Multi - task learning**:
- All \(\beta_j\) are simultaneously estimated by minimizing the loss function \(\frac{1}{2n}\|y - W\beta\|^2_2+\lambda_e\|\beta\|_1+\lambda_g\sum_{h = 1}^q\|b_h\|_2\), where \(\lambda_e\) and \(\lambda_g\) are tuning parameters.
3. **Debiasing method**:
- A piecewise debiasing method is proposed, which decomposes the inference of the entire \(\beta\) vector into piecewise inferences, avoiding the handling of cross - task dependencies.
- The projection technique is used to map the optimization problem from \(\mathbb{R}^{(p - 1)(q + 1)}\) to \(\mathbb{R}^n\), thereby significantly reducing the computational cost.
### Experimental results
- **Bias correction**: The effectiveness of the debiasing method is verified through simulation experiments. The debiased estimates are concentrated around the true parameter values and exhibit an approximately normal distribution.
- **Statistical inference**: The coverage rate of the constructed confidence intervals is close to the theoretical value, verifying the reliability of the method.
### Conclusion
By proposing efficient debiasing methods and projection techniques, this paper solves the statistical inference problems in high - dimensional Gaussian graphical regression models, providing a powerful tool for the analysis of complex biomedical data.