Hi, I have a vector which contains the cdf of a univariate distribution. The support of the rv is [0,1], so eg if the vector has 101 elements, the give the CDF at 0,0.01,...,1. The function is quite smooth. [1] I would like to calculate (approximate) the density. I tried fitting a splinefun and calling it with deriv=1, but occasionally this is negative. How can I ensure at least positivity? Thanks, Tamas [1] The cdf points are the result of a dynamic programming problem and is calculated numerically.