Difference between Two-Sample Means

This chapter is based on Tanizaki (1997), which is substantially revised. Nonparametric tests dealing with two samples include score tests (e.g., Wilcoxon rank sum test, normal score test, logistic score test and so on) and Fisher’s randomization test. Since in general the nonparametric tests require a large amount of computational burden, there are few studies on small sample properties although asymptotic properties from various aspects were studied in the past. In this chapter, the various nonparametric tests dealing with difference between two-sample means are compared with the t test, which is a conventional parametric test, through Monte Carlo experiments. Also, we consider testing structural changes as an application to the regression analysis.