Rob J Hyndman

Rob J Hyndman and Han Lin Shang

Journal of Com­pu­ta­tional and Graph­ical Stat­ist­ics (2010), 19(1), 29–45.

Abstract: We pro­pose new tools for visu­al­iz­ing large num­bers of func­tional data in the form of smooth curves or sur­faces. The pro­posed tools include func­tional ver­sions of the bag­plot and box­plot, and make use of the first two robust prin­cipal com­pon­ent scores, Tukey’s data depth and highest dens­ity regions.

By-​​products of our graph­ical dis­plays are out­lier detec­tion meth­ods for func­tional data. We com­pare these new out­lier detec­tion meth­ods with exist­ing meth­ods for detect­ing out­liers in func­tional data and show that our meth­ods are bet­ter able to identify the outliers.

Keywords: Highest dens­ity regions, Robust prin­cipal com­pon­ent ana­lysis, Ker­nel dens­ity estim­a­tion, Out­lier detec­tion, Tukey’s half­space depth.

Online paper

Work­ing paper

R pack­age