Title: Dynamic graphics for robust multivariate analysis in R
Authors: Emmanuele Sordini - Joint Research Centre of the European Commission (JRC) (Italy) [presenting]
Valentin Todorov - UNIDO (Austria)
Aldo Corbellini - Faculty of Economics - University of Parma (Italy)
Abstract: The monitoring of robust estimates computed over a range of key parameter values is a technique advocated in a number of recent articles. Through this approach the diagnostic tools of choice can be tuned in such a way that highly robust estimators which are as efficient as possible are obtained. The forward search for multivariate analysis is an algorithm for avoiding outliers by recursively constructing subsets of good observations and the underlying idea can be extended to many other techniques like S- and MM-estimates. To illustrate the forward search analysis, we start with a simple example and then analyze a real life data set. The analysis is conducted with the R package ``fsdaR'', which makes the analytical and graphical tools of the MATLAB FSDA library available to R users. The estimations are presented in monitoring plots of all $n$ squared Mahalanobis distances which can be combined with brushing to relate Mahalanobis distances to data points exhibited in scatterplot matrices. In this way, a straight relationship between statistical results and individual observations is established.