On certain exponential regularity for Gaussian processes (Q2463673): Difference between revisions

Let \(X=(X_j, j=1,\dots,m)\) be a zero-mean Gaussian vector with \(\sigma_j=\sqrt{E\;X_j^2}\). If \(f:\mathbb R^n\to \mathbb R\) is a Borel function, nondecreasing in each variable separately, and \[ f(x_1+s,\dots,x_m+s)\leq f(x_1,\dots,x_m)+s\;\forall x_i,\;\forall s>0, \] then \[ \exp\left[ -{\sigma_{\max}^2\over 2}+E(f(X)) \right] \leq E\left[ \exp\left( f(X_1-\sigma_1^2/2,\dots , X_m-\sigma_m^2/2) \right) \right] \leq\exp(E(f(X))). \] Based on these inequalities, the author derives lower and upper bounds for the ratio of option prices in Black-Scholes and Bachelier models.