Tianle Liu,Xiao-Li Meng,Natesh S. Pillai
Abstract:Recently, there has been a surge of interest in hypothesis testing methods for combining dependent studies without explicitly assessing their dependence. Among these, the Cauchy combination test (CCT) stands out for its approximate validity and power, leveraging a heavy-tail approximation insensitive to dependence. However, CCT is highly sensitive to large $p$-values and inverting it to construct confidence regions can result in regions lacking compactness, convexity, or connectivity. This article proposes a "heavily right" strategy by excluding the left half of the Cauchy distribution in the combination rule, retaining CCT's resilience to dependence while resolving its sensitivity to large $p$-values. Moreover, the Half-Cauchy combination as well as the harmonic mean approach guarantees bounded and convex confidence regions, distinguishing them as the only known combination tests with all such desirable properties. Efficient and accurate algorithms are introduced for implementing both methods. Additionally, we develop a divide-and-combine strategy for constructing confidence regions for high-dimensional mean estimation using the Half-Cauchy method, and empirically illustrate its advantages over the Hotelling $T^2$ approach. To demonstrate the practical utility of our Half-Cauchy approach, we apply it to network meta-analysis, constructing simultaneous confidence intervals for treatment effect comparisons across multiple clinical trials.
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is how to improve the existing combination test methods, especially the Cauchy combination test (CCT), in order to overcome its sensitivity to large p - values and the problems encountered when constructing confidence regions. Specifically:
1. **Sensitivity to large p - values**:
- CCT performs poorly when dealing with large p - values, which easily leads to numerical instability and loss of statistical power. For example, in genome - wide association studies, most p - values are close to 1, and only a few single - nucleotide polymorphisms (SNPs) are related to phenotypes, making CCT less suitable in such cases.
2. **Constructing reasonable confidence regions**:
- When constructing confidence regions by inverting the global test, CCT and other similar methods may produce non - convex or disconnected confidence regions, which is not ideal in practical applications. For example, when inverting CCT to obtain the confidence set of parameters, the acceptance region may be non - convex or even disconnected.
To solve these problems, the paper proposes a "heavily right strategy", that is, to improve the combination rule by excluding the left half of the Cauchy distribution, retaining the robustness of CCT to dependence, while solving its sensitivity to large p - values. Specifically, the paper introduces the Half - Cauchy combination test (HCCT) and the harmonic mean p - value (HMP) method. These two methods can ensure the boundedness and convexity of the confidence regions and perform well in practical applications such as high - dimensional mean estimation and network meta - analysis.
### Summary of main contributions
- **Improving combination test methods**: HCCT is proposed, which solves the problem of CCT's sensitivity to large p - values.
- **Constructing reasonable confidence regions**: HCCT and HMP can generate connected and convex confidence regions.
- **Efficient algorithms**: Efficient and exact algorithms for implementing HCCT and HMP are developed.
- **Practical applications**: The advantages of HCCT in high - dimensional mean estimation and network meta - analysis are demonstrated.
Through these improvements, the paper provides a more robust and effective combination test method, which is suitable for hypothesis testing and parameter estimation of dependent data.