logcondens: Computations Related to Univariate Log-Concave Density Estimation

Lutz Dümbgen, Kaspar Rufibach

Main Article Content


Maximum likelihood estimation of a log-concave density has attracted considerable attention over the last few years. Several algorithms have been proposed to estimate such a density. Two of those algorithms, an iterative convex minorant and an active set algorithm, are implemented in the R package logcondens. While these algorithms are discussed elsewhere, we describe in this paper the use of the logcondens package and discuss functions and datasets related to log-concave density estimation contained in the package. In particular, we provide functions to (1) compute the maximum likelihood estimate (MLE) as well as a smoothed log-concave density estimator derived from the MLE, (2) evaluate the estimated density, distribution and quantile functions at arbitrary points, (3) compute the characterizing functions of the MLE, (4) sample from the estimated distribution, and finally (5) perform a two-sample permutation test using a modified Kolmogorov-Smirnov test statistic. In addition, logcondens makes two datasets available that have been used to illustrate log-concave density estimation.

Article Details

Article Sidebar