The Faculty of Administration and Economics at the University of Kerbala has  discussed, via the electronic application, Zoom, the M,A. thesis entitled  “the use of some methods of estimating the approximate reliability function of contaminated data“.

 The thesis, submitted by M.A. student Muhammad Hussein Kadhim, aims at estimating the approximate reliability in the event that the data contains contaminated (anomalous) values ​​arising from the deviation of the original values ​​of the data from their primary distribution using the anomaly model of Dixit, which is used to find a common distribution of data in the event that it contains anomalies.

    The Thesis discusses the way of using the Ferrite distribution as the original distribution of data, the basic data has been contaminated with k of the values ​​following the exponential distribution as a first case and k of the values ​​follow the Whipple distribution as a second case, and the parameters of the Ferrite distribution were estimated using the greatest possible method and the placement method to estimate the distribution parameters in both cases,  Then, offset the estimates of those methods in the Ferrite distribution reliability function to obtain the approximate reliability of the distribution, and then compare the estimation methods using criteria and mean squares of the integral error MPALE.

 The thesis indicates that many statistical analyzes of a particular phenomenon, in which some observations happen to deviate or move away from the largest part of the observations that are present with it, and that deviant quarrels are called naming gay or pollutants.

     It  concludes that data contamination may be useful in some cases. In the case of real data, it is noted that deliberate downtime for the mammogram has led to an increase in the reliability of the device, and stressed the need to compare the method used in this thesis with the fortified methods that are used in the case of contaminated data and expansion in  use of distributions other than the Ferrite distribution in the case of contaminated data, as well as the use of distributions other than exponential distribution and Whipple distribution as contaminated distributions for the original distribution of data.