On Coupling of Brownian Bridges (Q4328532)

Let \(\{B(t),0<t<1\}\) be a standard Brownian bridge and let \(f: [0,1]\rightarrow \{+1,-1\}\) be a nonrandom measurable function. Denote by \(Y(t)=\int_0^t f(u)dB(u)\) and let \(X_+=\int_0^t 1_{\{f(u)=+1\}}dB(u)\), \(X_-=\int_0^t 1_{\{f(u)=-1\}}dB(u)\). The author deals with the following problem: Is there a Brownian bridge \(B^*\) such that \(|Y(t)|\geq|B^*(t)|\), \(0<t<1\) a.s.? The answer is positive, and the explicit form of \(B^*\) is presented, also a construction of Brownian bridge \(B^*\) such that \(\max_{0\leq t\leq 1}|B^*(t)|=\max_{0\leq t\leq 1}\{|X_+(t)|\bigvee |X_-(t)|\}\) is proposed. The idea of the proof is first to establish a discrete version of the results and then to use the weak convergence technique. As a consequence a useful inequality for the comparison of the tail distribution of \(\max_{0\leq t\leq 1}|B(t)|\) and \(\max_{0\leq t\leq 1}|Y(t)|\) is obtained, i.e. it is proved that \[ P\Bigl\{\max_{0\leq t\leq 1}|B(t)|\geq a\Bigl\}\leq P\Bigl\{\max_{0\leq t\leq 1}|Y(t)|\geq a\Bigr\}\leq P\Bigl\{\max_{0\leq t\leq 1}|B(t)|\geq a/2\Bigr\}. \] {}.