What should a researcher do when statistical analysis software terminates before completion with a message that the Hessian is not invertible? The standard textbook advice is to respecify the model, but this is another way of saying that the researcher should change the question being asked....