Statistical Graphics for Ranked Lists

• Activity Number: 80
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistical Graphics
• Author(s): Michael G. Schimek*+
• Companies: Medical University of Graz/Danube University Krems
• Keywords: data aggregation ; ranked list ; statistical graphics ; top-k list

Abstract

Assume we are confronted with ranked lists representing the same set of distinct objects produced by different assignment mechanisms (e.g. consumer preference). Even when the lists are short, it is almost impossible to identify features such as overlap of rank order without a visual aid because of the irregularities in such data. As a remedy we introduce a heatmap-like plot, but often we have to cope with very long lists (e.g. from microarray platforms). Then the top ranked objects are our concern under the assumption of a general decrease of the probability for consensus rankings with increasing distance from the top. Before we can use any graphical gadget, we have to estimate the points of rank order degeneration into noise for each pair of lists. The output of the procedure are, so-called, top-k lists we finally aggregate by means of graphical techniques and illustrate on real data.