Two Sample Test Besed On Rank Difference Sequence

In the statistical analysis methods,the non-parametric test is an important part.Among the general statistical inference,parametric tests and non-parametric tests constitute an important basis for content.Generally known in the form of general distribution,statistical inference methods to infer the population mean,variance and other test parameters are parameters.However,when analyzing the data,for various reasons,it assumed the form of the population distribution is unrealistic,so now not be able to use parametric test.Based on this case,the non-parametric test emerged.Generally,we do not know the overall distribution of forms or poorly understood,non-parametric tests using only sample data on the parameters of the overall distribution of statistical inference.This paper outlines the scope of the study and summarize some of the major research results of non-parametric tests,and focuses on determining whether the two sample population distribution of the same classic method,and then propose a new test thinking from another angle and construct a new the test statistic,the nature of the statistics were discussed.At the same time by the relative efficiency of tests to illustrate the applicability of difference sequence rank test,and finally show the actual situation by generating the data fitting.Content is roughly as follows:(A)rank sum test is seeking corresponding sample point rank in themixed sample and then the overall distribution is consistent with the size of the judgment based on rank.In this paper,the use of the mixed sample corresponding point difference between the stable rank of sample points uniformly dispersed in the case of the construction of new statistics.And distribution of statistics obtained,while in the large sample and small sample conditions given appropriate form distribution.(Ii)rank difference sequence of the specific nature of the test statistics were discussed and found a steady sequence of self-nature,and when the sample points are concentrated sequence is present decreasing trend,the sample points are concentrated at both ends of the sequences present growth trend.At the same time also pointed out that the existence of poor sequence rank test method limitations.(Iii)to test the applicability of rank difference sequence to explore,in the case of the overall symmetrical distribution relative test efficiency by t test,Wilcoxon rank sum test,rank difference sequence test under different circumstances of each applicable symmetrically.(Iv)the distribution of the generated data is too different in the case of the R software: symmetrical unimodal and bimodal symmetrical position is too different distribution,the same is too distribution.And divided into large samples and small sample of two states,the data to the operator compare Wilcoxon rank sum test and rank difference sequence test their effectiveness.