STATISTICAL DECISIONS IN PRESENCE OF IMPRECISELY REPORTED ATTRIBUTE DATA

摘要

The paper presents a new methodology for making statistical decisions when data is reported in an imprecise way. Such situations happen very frequently when quality features are evaluated by humans. We have demonstrated that traditional models based either on the multinomial distribution or on predefined linguistic variables may be insufficient for making correct decisions. Our model, which uses the concept of the possibility distribution, allows to separate stochastic randomness from fuzzy imprecision, and provides a decision - maker with more information about the phenomenon of interest.