Statistical Analysis About Some Problems In PSC

PSC is a language test which is of great importance. The current regulations of PSC are based on many years of testing experience, and the study for it was mostly stayed on qualitative analysis/And more, PSC itself is a subjective test. All of those made the current regulations lack of scientificalness and rationality, and can't offer a method to control the quality of the test.To build scientifical and rational PSC regulations, we must find out the faults in current regulations and improve them by compact and quantitative methods.This paper found two faults in current regulations. The first, it will cause decrease of reliability of PSC to take granted all the candidates' score without distinguishing the examiners with low reliability from the others. The second, the range of re-test is too small. The candidates who are mis-judged to the lower gradefrom the higher one are not re-tested, and the same as the candidates who are not of the first grade.This paper dealt with the two faults by statistical methods. For the first one, we offer a way to distinguish the examiners with low reliability from the others, and correct the scores judged by such examiners. For the second one, the methods of interval estimate are used to find the intervals where the mis-judges of grade are most probably happen. Such intervals are regarded as the re-test interval, and so the new re-test regulations are built.