The two-parameter Brownian bridge: Kolmogorov inequalities and upper and lower bounds for the distribution of the maximum (Q789108)

Let \(\beta(t,s)\) be a 2-parameter Wiener process, \(M(x,y)=\sup \{\beta(t,s),t\in [0,x],s\in [0,y]\}\) and \(\phi(t)\) be the standard normal density. The authors investigate the probability density \[ p(x,y,z,u)=p(t,z,y)=(\partial /\partial z)P\{M(x,y)<u,\quad \beta(x,y)>u- z\} \] where \(t=xy\) and correct their earlier results [the first author, Z. Wahrscheinlichkeitstheor. Verw. Geb. 48, 285-292 (1979; Zbl 0388.60080) and the present authors, C. R. Acad. Sci., Paris, Sér. A 289, 453-454 (1979; Zbl 0438.60064)]. If D denotes the differential heat operator \(D=frac{1}{2}\partial^ 2/\partial z^ 2-\partial /\partial t\) then p(t,z,u) satisfies the inequalities \(Dp\geq 0\) and \(DtDp\geq 0\) as a distribution. With the help of \textit{V. Goodman's} ideas [Ann. Probab. 4, 977-982 (1976; Zbl 0344.60048)] the upper bound for the density \(p(t,z,u)\leq p_ 0(t,z)- 2uzt^{-3/2}\phi(t^{-1/2}(u+z))\) where \(p_ 0(t,z,u)=t^{- 1/2}\{\phi(t^{-1/2}(u-z)-\phi(t^{-1/2}(u+z))\}\) is obtained. Also, the lower bound \(p(t,z,u)\geq \tilde p(t,z,u)\) where \[ \tilde{p}(t,z,u) = p_ 0(t,z,u) - \int^{t}_{0} (\log t-\log s)(t-s)^{-3/2} z \phi(z(t-s)^{-1/2}) 2u^2 s^{-3/2} \phi(us^{-1/2})ds \] is given. These estimates lead to the upper and lower bounds for the yet unknown distribution of the maximum of the multiparameter Wiener process; see also: \textit{C. Park}, \textit{D. L. Skoug} [Pac. J. Math. 78, 455-466 (1978; Zbl 0394.60079)]; \textit{V. Piterbarg} and \textit{V. Futulov}, ''Probability and statistical methods of exploration.'' in A. N. Kolmogorov and I. G. Zhurbenko (eds.), Veroyatnostno-statisticheskie metody issledovaniya, 124-143 (1983). Some results of numerical calculations are given.