We consider the problem of estimating a density function that is known in advance to be convex. The maximum likelihood estimator is then the solution of a linearly constrained convex minimization problem. This problem turns out to be numerically difficult. We show that interior point algorithms perform well on this class of optimization problems, though for large samples, numerical difficulties are still encountered. To eliminate those difficulties, we propose a clustering scheme that is reasonable from a statistical point of view. We display results for problems with up to 40000 observations. We also give a typical picture of the estimated density: a piece wise linear function, with very few pieces only.
Report 95-49, Faculty of Technical Mathematics and Computer Science, Delft University of Technology, Delft, 1995.