CMStatistics 2015: Start Registration
View Submission - CMStatistics
B1310
Topic: Contributed on Statistical methods for big data and antifraud analysis Title: Analysing large datasets with the forward search in SAS Authors:  Francesca Torti - European Commission (Italy) [presenting]
Marco Riani - University of Parma (Italy)
Abstract: The application of robust methods to international trade data may present serious scalability problems because the sample size $n$ typically ranges from few tens to several hundreds of thousands units. This is particularly true for the Forward Search (FS), which needs to build a series of subsets of size increasing from few units (say $v$, i.e. the number of data variables) to $n$ units. For this reason,we have implemented a SAS package for the FS that complements the official MATLAB implementation FSDA. We illustrate the main features of the new SAS FS package on a number of challenging international trade datasets provided by the customs services of some Member States. The illustration will also cover the typical interactive graphical tools for exploratory data analysis offered by FSDA.