We are concerned with the likelihood ratio tests in the $p_0$ model for testing degree heterogeneity in directed networks. It is an exponential family distribution on directed graphs with the out-degree sequence and the in-degree sequence as naturally sufficient statistics. For two growing dimensional null hypotheses: a specified null $H_{0}: \theta_{i}=\theta_{i}^{0}$ for $i=1,\ldots,r$ and a homogenous null $H_{0}: \theta_{1}=\cdots=\theta_{r}$, we reveal high dimensional Wilks' phenomena that the normalized log-likelihood ratio statistic, $[2\{\ell(\widehat{\bs\theta})-\ell(\widehat{\bs\theta}^{0})\}-r]/(2r)^{1/2}$, converges in distribution to a standard normal distribution as $r\rightarrow \infty$. Here, $\ell( \bs{\theta})$ is the log-likelihood function, $\widehat{\bs{\theta}}$ is the unrestricted maximum likelihood estimator (MLE) of $\bs\theta$, and $\widehat{\bs{\theta}}^0$ is the restricted MLE for $\bs\theta$ under the null $H_{0}$. For the homogenous null $H_0: \theta_1=\cdots=\theta_r$ with a fixed $r$, we establish the Wilks-type theorem that $2\{\ell(\widehat{\bs{\theta}}) - \ell(\widehat{\bs{\theta}}^0)\}$ converges in distribution to a chi-square distribution with $r-1$ degrees of freedom as $n\rightarrow \infty$, not depending on the nuisance parameters. These results extend a recent work by \cite{yan2023likelihood} to directed graphs. Simulation studies and real data analyses illustrate the theoretical results.
翻译:暂无翻译