Maximum penalized likelihood estimation. Vol. 1: Density estimation (Q5940556): Difference between revisions

This is the first volume of a two-volume text on theory and practice of maximum likelihood estimation. Whereas Volume II is concerned with indirect estimation problems, this volume deals with density estimation, mainly with the nonparametric approach. The authors introduce the maximum (penalized) likelihood method for estimating a density function, and consider existence and uniqueness of the proposed estimators and derive almost sure convergence rates for the \(L_1\)-error. Estimation under constraints, such as unimodality and log-concavity, is studied in more detail. Furthermore, data-driven smoothing parameter selection methods are discussed. The reader will gain insight into some of the generally applicable technical tools from probability theory (discrete parameter martingales) and applied mathematics (boundary value problems). The theoretic results are completed by ``in action chapters'', where computational issues are discussed and the small sample behavior of the estimators is demonstrated using simulated and real data. Since the maximum penalized likelihood estimators are defined as solutions to convex minimization problems, convexity and optimization are treated in a special part. The text is intended for graduate students in statistics, applied mathematics, and operations research, as well as for researchers and practitioners in the field.