Tail probabilities of the maxima of Gaussian random fields (Q2365735)

Let \(\{Z(t)\), \(t\in I\}\) be a nonsingular Gaussian random field with zero mean and unit variance where \(I\) is a \(d\)-dimensional index set. In several statistical inference problems, it is important to calculate \[ P[\max\{Z(t),\;t\in I\}\geq{\mathfrak z}] \tag{*} \] as \({\mathfrak z}\to\infty\) [cf. \textit{J. Sun}, Biometrika 78, No. 4, 759-769 (1991; Zbl 0753.62067)]. A one term approximation formula for \((*)\) is given \textit{R. J. Adler} [The geometry of random fields (1981; Zbl 0478.60059), p. 160] if the Gaussian random field is homogeneous. This approximation is inaccurate when \(d\) is large and Gaussian random fields derived in statistical problems are not homogeneous in general. Here the author obtains a higher-order approximation formula for a wider class of Gaussian random fields. The author uses a ``tube method'' by using the \textit{H. Weyl}'s formula [Am. J. Math. 61, 461-472 (1939; Zbl 0021.35503)].