In $$\Lambda$$, the likelihood in the denominator is $$L(\widehat{\theta}_s|x)$$, where $$s$$ corresponds to the saturated model: a model that has one parameter for each observation $$x_i$$.