Likelihood principle — proposition that all statistical evidence in a sample that are relevant to model parameters is contained in the likelihood function.