FSDA (Forward Search for Data Analysis) extends MATLAB and the Statistics Toolbox for the robust analysis of data sets affected by different sources of heterogeneity. It includes the main traditional robust multivariate and regression techniques (LMS, LTS, MCD, MVE, MM and S estimation) and the Forward Search approach. Tools are available for robust data transformation and robust model selection. The exploratory plots produced by the statistical functions are enriched with flexible dynamic interaction options, based on brushing, linking, annotation and filtering paradigms. The documentation is extensive and fully integrated in the classical documentation system and style of MATLAB. Documentation pages include code fragments with examples taken from the robust statistics literature. The corresponding datasets (about 50, currently) are available in textual (.txt) and binary (.mat) formats. Code for any function is open and extensible.
Alternative download link: