Rejoinder to “On the robustness of latent class models for diagnostic testing with no gold standard”