Some approaches to increase the robustness of likelihood functions in their application for estimation of parameters of stock assessment models are considered.