This Monday, during my talk on quantile regressions (at the Montreal R-meeting), we've seen how those nice graphs could be interpreted, with the evolution of the slope of the linear regression, as a function of the probability level. One illustrati...