Hypothesis testing for homogenous of nodes in $β$-models

Kang Fu,Jianwei Hu,Meng Sun

2024-03-10

Abstract:The $\beta$-model has been extensively utilized to model degree heterogeneity in networks, wherein each node is assigned a unique parameter. In this article, we consider the hypothesis testing problem that two nodes $i$ and $j$ of a $\beta$-model have the same node parameter. We prove that the null distribution of the proposed statistic converges in distribution to the standard normal distribution. Further, we investigate the homogeneous test for $\beta$-model by combining individual $p$-values to aggregate small effects of multiple tests. Both simulation studies and real-world data examples indicate that the proposed method works well.

Statistics Theory

What problem does this paper attempt to address?

This paper aims to address the hypothesis testing problem of node parameter homogeneity in the β model. Specifically, the authors propose a statistic to test whether two nodes have the same parameter value in the β model. Additionally, the paper explores how to conduct an overall homogeneity test for the β model by combining the p-values of multiple tests.

### Main Issues
1. **Node Parameter Equality Test**:
- Hypothesis testing problem: For any two nodes \(i\) and \(j\), test whether their parameters are equal, i.e., \(H_0: \beta_i = \beta_j\) versus \(H_1: \beta_i \neq \beta_j\).
- The authors prove that under the null hypothesis, the proposed statistic's limiting distribution is a standard normal distribution, and under the alternative hypothesis, the statistic's limiting distribution is a normal distribution with a different mean.

2. **Overall Homogeneity Test**:
- Hypothesis testing problem: Test whether all nodes have the same parameter, i.e., \(H'_0: \beta_1 = \beta_2 = \cdots = \beta_n\) versus \(H'_1: \beta_i \neq \beta_j\) for at least one pair \(i, j\).
- The authors propose a method based on combining p-values to address this problem, specifically using the Cauchy combination method, which remains effective under dependency structures.

### Methods and Results
- **Node Parameter Equality Test**:
- The authors propose a statistic \(U_{ij} = \frac{\hat{\beta}_i - \hat{\beta}_j}{\sqrt{\hat{v}_{ii}^{-1} + \hat{v}_{jj}^{-1}}}\) and prove that its limiting distribution under the null hypothesis is a standard normal distribution.
- Through simulation studies, they verify that this statistic performs well in finite samples.

- **Overall Homogeneity Test**:
- The authors use the Cauchy combination method to combine the test results of multiple node pairs, forming an overall test statistic \(T_n\).
- Simulation results show that this method has high test power while controlling the type I error rate.

### Application Example
- The authors apply the proposed method to a real network dataset—the Chesapeake Bay food web dataset. The results indicate that the network is homogeneous with high probability.

### Conclusion
- This paper proposes a new statistical method to test the equality and overall homogeneity of node parameters in the β model. The effectiveness and efficiency of the method are validated through theoretical analysis and simulation studies. Future research could consider how to improve test power when parameter differences are small and extend from single-sample to multi-sample scenarios.