Estimating a correlation coefficient when one variable is not directly observed

The problem is to estimate p the true correlation between two variables x and y when y and u=f(x) (but not x) are measured where f is a monotone function. The monotone correlation coefficient between y and u is defined and shown to be a (strongly) consistent estimator of p. For large samples approximate lower and upper bounds on the sample correlation between x and y are presented. Finally, an application of these results is displayed.