Robust Statistics

Robust statistics is a scientific field, mostly devoted to analyse data with methods which are said robust, so that results are not affected by atypical or influential values, often referred to as outliers.

Carbon dating of the Shroud case.

FSDA Toolbox

The FSDA Toolbox is jointly developed with the Joint Research Centre of the European Commission. The FSDA is a MATLAB Toolbox which enhance the statistics Toolbox and provide reliable support to robust and efficient data analysis for complex data.
The scientific publication "The FSDA Matlab toolbox: An integrated framework to assess and apply robust methods to complex datasets" was awarded as the best contribution during the MATLAB EXPO 2016, held in Milan in June 2016.


The Mission of the Center is to develop new methods and disseminate propertied of robust statistics in many applied fields. The Ro.S.A. has contributed to give support for the solution of data-based problems in public and private industries.

Robust Analysis of Big Data

Data are a new fuel; their collection into growing databases is becoming an endless source of information for many brands. Robust statistics can be groundbreaking compared to obsolete traditional methods.

Forward Search

The Forward Search is a novel yet powerful method whose development dates back of about 20 years from Ro.S.A.'s founders. It is a general method which encapsulate a dynamic data-driven set of recursions making simple the identification of influential observations and unmask hidden structure in the data, so that robust techniques can be coherently developed.

The Robust Statistics Academy is a Research Center set in the University of Parma. It has been built by a group of statisticians, with long-term experience. Their scientific works is based on the modern usage of statistical models, finely tuned to be robust.

The Robust Statistics Academy has the aim of disseminating the properties of robust statistical methods and share their results with other Research Centres and private companies.

The robust approach gives the opportunity to easily cope with extreme values and outliers. It is a fundamental bag, with many tools, usable with Big Data.

Why robust statistics?


1 Barnett and Lewis: "We will define an outlier in a set of data to be an observation (or subset of observations) that seems inconsistent with the rest of that data set." The Robust Statistics reduces incorrect measurement risks generated by outliers.

Influential observations

2In some applied fields, often are the so-called "outliers" which are of interest (fraud detection is an example). Sometimes, additionally, observations might cluster in such a way to become interesting for their own pattern, which with robust methods can be properly disentangled from the bulk of the data.

Big Data

3It almost unavoidable, when analyzing big data, to avoid errors or outliers. With Robust Statistics the final results are not affected by those.


4The MATLABĀ® Toolbox FSDA "Flexible Statistics for Data Analysis Toolbox", developed by the Center, is a powerful tool to be used in data analysis; it us maintained and updated regularly with new tools.