A MATLAB Toolbox for Robust Statistics: Use in Policy Research

The Joint Research Centre of the European Commission and the University of Parma have developed a MATLAB^® toolbox, Flexible Statistics and Data Analysis (FSDA), for supporting relevant European Union policies and for conducting statistical research. The user community of FSDA is now wide and heterogeneous, and benefits from extensive, technically precise, and regularly updated documentation. The talk illustrates the use of FSDA in applications of compelling importance for the European Union policies.

FSDA extends MATLAB and the Statistics and Machine Learning Toolbox™ to support a robust and efficient analysis of complex data sets affected by different sources of heterogeneity, ensuring an output unaffected by anomalies in the provided data. FSDA provides statisticians, engineers, scientists, researchers, and financial analysts with a comprehensive set of tools to assess and understand their data, including interactive features for analyzing and modeling data, and learning and teaching statistics. More specifically, the robust statistical functions inside FSDA cover regression analysis, multivariate analysis, data transformations in regression and multivariate applications, model selection, clustering, correspondence analysis, time series analysis, and interactive statistical visualization. 

FSDA is developed for wide applicability. For its capacity to address problems focusing on anomalies in the data, it is currently used in applications such as anti-fraud, detection of computer network intrusions, e-commerce and credit card frauds, customer and market segmentation, detection of spurious signals in data acquisition systems, in chemometrics (a wide field covering biochemistry, medicine, biology, and chemical engineering), in issues related to the production of official statistics (e.g., imputation and data quality checks), and so on. 

FSDA can be freely downloaded from Flexible Statistics and Data Analysis Toolbox and its documentation can be previewed or consulted directly in the same website. All FSDA functions are written in the open MATLAB language to facilitate the understanding of the algorithms implemented and encourage the user to enrich the toolbox with new functions. It is customary for a MATLAB user to document the new functions only in the head of the .m file. Only rarely is the user prepared to duplicate the effort and work on the corresponding html documentation file. This is understandable, since the complete integration of new html files in the standard MATLAB documentation system is not facilitated by built-in tools: FSDA includes parsing functions that translate automatically the ascii text in the .m files into well-formatted html files in the standard MATLAB documentation style. 

Joint work with Marco Riani and Aldo Corbellini of University of Parma, Italy.