On the equivalence of the tube and Euler characteristic methods for the distribution of the maximum of Gaussian fields over piecewise smooth domains (Q1872403): Difference between revisions

Let \(M\) be a closed subset of the unit sphere \(S^{n-1}\) in \(R^n\). Consider a Gaussian random field \(\{Z(u)\mid u=(u_1,\dots ,u_n)'\in M\}\) with a finite Karhunen-Loeve expansion of the form \(Z(u)=\sum^n_{i=1}u_iz_i,\) where \(z_i, i=1,\dots ,n,\) are independent standard normal variables. The asymptotic behavior of the upper tail probability \(P_x:=P(\max_{u\in M}Z(u)\geq x) \) is studied. In order to derive the asymptotic expansion of the upper tail probability \(P_x\) for \(Z(u)\), two methods are currently available. One is the tube method devoted by \textit{J. Sun} [Ann. Probab. 21, 34-71 (1993; Zbl 0772.60038)]. The other method is the Euler characteristic method developed by \textit{R. J. Adler} [``The geometry of random fields'' (1981; Zbl 0478.60059)] and \textit{K. J. Worsley} [Ann. Stat. 23, 640-669 (1995; Zbl 0898.62120) and Adv. Appl. Probab. 27, 943-959 (1995; Zbl 0836.60043)]. In the paper, under the very general assumption that \(M\) is a manifold with a piecewise smooth boundary it is proven that the tube and Euler characteristic methods are equivalent. In order to show the equivalence a version of the Morse theorem for a manifold with a piecewise smooth boundary is proved.