Demetriou describes an implementation in Fortran 77 of Demetriou and Powell’s algorithm to fit data contaminated with random errors to a convex function. The technique first approximates a solution, then uses quadratic programming with iterative refinement and automatic knot placement on B-splines to find the solution. The technique is useful for large data sets where convexity is assured, and the author cites economic models as a suitable application area. The paper is best seen as an addendum to the original paper [1], describing the practical implementation of the method.