Identifiability of probabilistic models, with examples from knowledge structure theory