PN-II-RU-TE-2012-3-0007
FLEXIBLE STATISTICAL MODELING AND ROBUST ANALYSIS OF FINANCIAL DATA
Financed by CNCS-UEFISCDI
Contract 30/29.04.2013
Abstract: Outliers in financial data are a frequently occurring phenomenon and can have an adverse influence on the performance of classical statistical methods for analyzing these data. In an information theoretic framework, the project will develop new robust statistical methods for estimation and testing, considering parametric models, regression models and moment condition models. Robust model selection criteria will be constructed as well. The main approach will be based on the use of divergences between probability measures with accent on duality or decomposability arguments allowing defining robust statistical criteria. Theoretical and empirical properties of the estimators and tests will be analyzed. The new methods will be applied in asset allocation models, in estimating financial risks, as well as in other financial models. The performance of these methods will be investigated for both simulated and real financial data.