18
0

Likelihood ratio tests in random graph models with increasing dimensions

Abstract

We explore the Wilks phenomena in two random graph models: the β\beta-model and the Bradley-Terry model. For two increasing dimensional null hypotheses, including a specified null H0:βi=βi0H_0: \beta_i=\beta_i^0 for i=1,,ri=1,\ldots, r and a homogenous null H0:β1==βrH_0: \beta_1=\cdots=\beta_r, we reveal high dimensional Wilks' phenomena that the normalized log-likelihood ratio statistic, [2{(β^)(β^0)}r]/(2r)1/2[2\{\ell(\widehat{\mathbf{\beta}}) - \ell(\widehat{\mathbf{\beta}}^0)\} -r]/(2r)^{1/2}, converges in distribution to the standard normal distribution as rr goes to infinity. Here, (β)\ell( \mathbf{\beta}) is the log-likelihood function on the model parameter β=(β1,,βn)\mathbf{\beta}=(\beta_1, \ldots, \beta_n)^\top, β^\widehat{\mathbf{\beta}} is its maximum likelihood estimator (MLE) under the full parameter space, and β^0\widehat{\mathbf{\beta}}^0 is the restricted MLE under the null parameter space. For the homogenous null with a fixed rr, we establish Wilks-type theorems that 2{(β^)(β^0)}2\{\ell(\widehat{\mathbf{\beta}}) - \ell(\widehat{\mathbf{\beta}}^0)\} converges in distribution to a chi-square distribution with r1r-1 degrees of freedom, as the total number of parameters, nn, goes to infinity. When testing the fixed dimensional specified null, we find that its asymptotic null distribution is a chi-square distribution in the β\beta-model. However, unexpectedly, this is not true in the Bradley-Terry model. By developing several novel technical methods for asymptotic expansion, we explore Wilks type results in a principled manner; these principled methods should be applicable to a class of random graph models beyond the β\beta-model and the Bradley-Terry model. Simulation studies and real network data applications further demonstrate the theoretical results.

View on arXiv
Comments on this paper