Choice of the hypothesis matrix for using the Wald-type-statistic

A widely used formulation for null hypotheses in the analysis of multivariate d-dimensional data is H-0 : H theta = y with H is an element of R-mxd, theta is an element of R-d and y is an element of R-m, where m <= d Here the unknown parameter vector theta can, for example, be the expectation vector mu, a vector beta containing regression coefficients or a quantile vector q. Also, the vector of nonparametric relative effects p or an upper triangular vectorized covariance matrix v are useful choices. However, even without multiplying the hypothesis with a scalar gamma not equal 0, there is a multitude of possibilities to formulate the same null hypothesis with different hypothesis matrices H and corresponding vectors y. Although it is a well-known fact that in case of y = 0 there exists a unique projection matrix P with H theta = 0 double left right arrow P theta = 0, for y not equal 0 such a projection matrix does not necessarily exist. Moreover, such hypotheses are often investigated using a quadratic form as the test statistic and the corresponding projection matrices frequently contain zero rows; so, they are not even efficient from a computational point of view. In this manuscript, we show that for the Wald-type-statistic (WTS), which is one of the most frequently used quadratic forms, the choice of the concrete hypothesis matrix does not affect the test decision. Moreover, some simulations are conducted to investigate the possible influence of the hypothesis matrix on the computation time.