Research On Confidence Discrimination Based On Reference Domain

Gr(?)sbeck and Saris first proposed the reference value to observe the fluctuation of blood analyte concentration;however,the reference domain is the interval constructed according to the sample,and the purpose is to contain(100?)% of the value of the population of interest.It is used to analyze the observation value and judge whether a certain index is normal under a certain confidence level in clinic.For example,in medicine,if the observation value from the population exceeds this interval,it will be classified as atypical and thus may need further diagnosis.Since the reference domain is constructed based on random samples,the event ”the reference domain contains(100?)% of the population” is random,and what we expect is that the event can have a large probability of occurrence.Firstly,this paper constructs and discusses the reference domain from normal and nonparametric aspects,and the two kinds of reference domain are compared by software drawing.It is found that the probability of ? prediction domain containing less than(100?)% of the population is very high,while the tolerance domain contains(100?)% of the population under the pre-given large confidence level ?,so it is considered that the tolerance domain is more suitable as the reference domain than the ? prediction domain.Secondly,by adjusting the parameters,we can observe the sensitivity of the reference domain to the parameters,such as changing the sample size n,we can see that ? prediction domain is less affected by n,while the tolerance domain and nonparametric reference domain will narrow with the increase of n,indicating that when n is large enough,the construction of reference domain will be better.Finally,it is applied to the case of postgraduate entrance examination,that is,the results of the first examination of a postgraduate entrance examination student are known,and the second examination is not held.On this premise,the reference domain of the examinee's initial examination results is calculated with a certain confidence level,and the admission rate is calculated in three cases according to the reference domain of the results.It can also be seen from the paper that equal-tailed(?,?)tolerance domain and symmetric nonparametric(?,?)tolerance domain have achieved good results in empirical research.