Imputation and outlier detection in banking datasets

Data from the banking balance sheets can be used to analyse the financial stability of the banking sector. Occasionally, it may occur that some data values are either incorrect or missing, which would have an important effect on the results of the analyses. Thus, incorrect values should be detected and removed or corrected, while missing values should be imputed. This contribution addresses the two problems using a robust data analysis approach, known as Forward Search. In particular, the Forward Search is used to address the presence of high data collinearity, which may give rise to many irrelevant outliers. In recent years a MATLAB toolbox, the Forward Search for Data Analysis (FSDA), has been applied to similar problems in official statistics. The contribution extends the application to the banking sector.

PERROTTA Domenico;  ARSENIS Spyros;  PAGANO Andrea; 

2014-08-25

CLEUP

JRC71523

978-88-6129-882-8,   

http://www.sis-statistica.it/index.php?area=main&module=contents&contentid=527,    http://new.sis-statistica.org/wp-content/uploads/2013/09/RS12-Imputation-and-outlier-detection-in-banking.pdf,    https://publications.jrc.ec.europa.eu/repository/handle/JRC71523,