If a test statistic $T$ follows a distribution $f(t; \theta)$ and the Likelihood Ratio $\Lambda = \frac{L(\theta_0)}{L(\hat{\theta})}$, what does the theorem of Wilks state about $-2 \ln \Lambda$ as $n \to \infty$?