CMStatistics 2016: Start Registration
View Submission - CMStatistics
B0282
Title: Choosing graphics for exploring data: Scagnostics and interaction Authors:  Katrin Grimm - University of Augsburg (Germany) [presenting]
Abstract: Choosing the right graphics for understanding and analyzing new data sets with a huge number of variables can be time intensive and challenging. To get first insights into important variables and interesting structures a small number of well chosen criteria and the right interactive tools can provide really helpful support. This talk concentrates on scagnostics, which are measures for characterizing scatterplots. Scagnostics (from scatterplot and diagnostics) were first mentioned by Paul and John Tukey in the 1980s. Concrete measures were previously proposed and implemented in the R-package scagnostics. Specifically, clustering plots using nine scagnostics were suggested to identify both unusual plots (outliers) and frequently occurring plots (summarised by exemplars). Providing flexible interactive interfaces for the criteria gives the user more control of the analysis. The aim is to introduce an interactive apps programmed with the R-package shiny and shows the functionality using a dataset of the 2013 German election.