Η επίδραση της λογοκρισίας στα δεδομένα στην εκτίμηση των παραμέτρων στις κατανομές απώλειας
The effects of censoring at the data at the estimation of parameters for the loss distributions
SubjectΕκτίμηση, Θεωρία της
Censoring in a sample can have a significant impact on parameter estimation. This paper studied the censoring type I, type II and progressive censoring and the effect they have on the efficiency of the maximum likelihood estimator for loss distributions. Simulated samples will be used from different loss distributions. The first chapter is a brief reference to basic concepts such as loss distributions, the Fisher information and maximum likelihood estimators. The second chapter refers to censoring and its types (censored type I, type II and random censoring). In third, fourth and fifth chapter presents an analysis of Weibull, Pareto and Gamma distribution used in this work. The individual chapters are presented also results from the application of the distributions in Mathematica program. The algorithms listed in the Annex. Finally, the sixth chapter presents the goodness of fit in Weibull, Pareto and Gamma distribution samples from which better adaptation presented by Pareto. Then the distribution is controlled by the effect of censoring degree in various statistical characteristics.