1. Because the proposed method executes without a model, its application is limited to explain observed data but to predict the response probability for an examinee on an item in the future.
2. From the results, the precision of classification could be seriously affected by either large guess or slip. That is to say, the method can work perfectly as well as the parametric approach only when both guess and slip parameters equal to zero. Empirically the requirement is too harsh to be fulfilled, suggesting proposed method is very likely to report less precise classification.
3. Similarly, other noise such as DIF can decrease the efficiency of this method.