The verification of the multivariate normal distribution hypothesis in a one-sample model with the method of elimination of disturbing parameters

Abstract

In many statistical tasks a necessity of testing multivariate normality arises. In constructing multivariate normality tests there is a necessity of estimating unknown parameters ц and £ from a given sample. The parameters are regarded as disturbing parameters. The paper deals with some methods, by means of which unknown disturbing parameters are eliminated when the multivariate normality tests are applied. In particular, the following methods are stressed: randomization method, reduction methods and conditional interval probability transformation method.