Kijko, Andrzej2025-12-082025-12-082019-042019-02*A2019http://hdl.handle.net/2263/107113Dissertation (MSc (Actuarial Science))--University of Pretoria, 2019.Systems and processes may fail and employees can engage in fraudulent ac-tivities that can go unnoticed for a very long time and the resulting losses can be very high and catastrophic to an institution. Setting a minimum threshold or a level of completeness will not guarantee that all losses above this point will be reported. In order to model operational risk data, a method that does not depend on the level of completeness is suggested. This can be done by introducing a de-tection probability that is combined with the underlying loss distribution to give a 3-parameter gamma distribution and fitted to a simulated dataset. It is found that the methodology is able to accurately estimate parameters when the data is incomplete.en© 2024 University of Pretoria. All rights reserved. The copyright in this work vests in the University of Pretoria. No part of this work may be reproduced or transmitted in any form or by any means, without the prior written permission of the University of Pretoria.UCTDSustainable Development Goals (SDGs)Loss data analysisOperational riskLevel of completionFrau detectionGutenburg-Richter b-valueDissertationu12235947N/A