Optimization cannot proceed because the initial value of the negative log-likelihood is negative infinity. This usually occurs when the model perfectly fits the data with the specified initial parameter values.