fsdaSAS brings to the SAS community some popular robust methods. The peculiarity of the package is that the methods are revisited and embedded within the so-called monitoring framework, where the properties of the fitted model are checked over a series of breakdown point values or, for the Forward Search (FS) method that inspired the approach, over a series of subsets of increasing size. The extension to the monitoring of popular estimators such as the LTS and LMS is a particularly powerful new feature and a novelty in the statistical literature. In addition, in this package the FS is extended with a batch option for analysing very large datasets; on the opposite side, the package also modifies the standard LTS and LMS IML functions by introducing the small sample correction factor of Pison and by increasing the range of values of the trimming parameter h in LTS. In fsdaSAS, the interpretation of the monitored results becomes easier with the ehnanced graphics provided by SAS IML Studio. The package includes these main functions:

• Monitoring.sx for monitoring a number of traditional robust multivariate and regression estimators (S, MM, LTS and LMS), already present in SAS, for specific choices of breakdown point or efficiency.
• FSR.sx and FSM.sx, which implement the FS approach to detect outliers, respectively, in regression and in multivariate data;
• FSRfan.sx and FSMfan.sx for identifying the best transformation parameter for the Box–Cox transformation in regression and multivariate analysis ([31], Chapter 4);
• FSM.sx, the multivariate counterparts of FSR, and FSMfan.sx for multivariate transformations;
• FSRms.sx for choosing the best model in regression.
• FSRMultipleStart.sx and FSMmultiplestart.sx for identifying observations that are divided into groups either of regression models or of multivariate normal clusters.