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 \(x_i\).