The author adopted the concept of classical difficulty to set up a optimization problem—the difference of model-based frequency and the empirical frequency. Then a series of simulation studies were conducted to realize the need of sample size for stable estimation.
1. The concept is simple but hard to follow what author described and replicate their study.
2. The attributes have been marginalized in formula. What is the meaning of expectation in EM?
3. There is only one criterion to optimize. It could include many other criteria such as one proposed by de la Torre (2008) to achieve simultaneously.