Error bounds for asymptotic expansions of the maximums of the multivariate t- and F-variables with common denominator (Q920497)

Let \(X=(X_ 1,...,X_ p)\) be a scale mixture of a p-dimensional random vector \(Z=(Z_ 1,...,Z_ p)\) with scale factor \(\sigma >0\), i.e. \(X=\sigma Z\) where Z and \(\sigma\) are independent. The author is concerned with asymptotic expansions of the distribution function of Max \(\{\) \(X_ j\}=\max \{X_ 1,...,X_ p\}\). The purpose of this paper is to give a unified derivation of the asymptotic expansions as well as their error bounds. First, two types of asymptotic approximations for the distribution function of X and their bounds are given. These results are applied to study the distribution of Max \(\{\) \(X_ j\}\) in the case when \(Z_ 1,...,Z_ p\) are independent and identically distributed. Finally, the author obtains asymptotic expansions of the distributions of Max \(\{\) \(X_ j\}\) and their error bounds in two important special cases: (i) \(Z_ 1,...,Z_ p\) are independent, normally distributed random variables (r.v.'s) and \(\sigma =(\chi^ 2_ n/n)^{1/2}\) where \(\chi^ 2_ n\) is the r.v. which has \(\chi^ 2\)-distribution with n degrees of freedom; (ii) \(Z_ 1,...,Z_ p\) are independent gamma- distributed r.v.'s and \(\sigma =\chi^ 2_ n/n\).