Optimizing an objective function under a bivariate probability model