Scagnostics

Scatterplot matrix of the scagnostics measures for the 91 scatterplots of the variables of the Boston Housing data set

Scagnostics (scatterplot diagnostics) is a series of measures that characterize certain properties of a point cloud in a scatter plot. The term and idea was coined by John Tukey and Paul Tukey, though they didn't publish it; later it was elaborated by Wilkinson, Anand, and Grossman. The following nine dimensions are considered:[1][2]

  1. For the outliers in the data:
    1. outlying
  2. For the density of data points:
    1. skewed
    2. clumpy
    3. sparse
    4. striated
  3. For the shape of the point cloud:
    1. convex
    2. skinny
    3. stringy
  4. For trends in the data:
    1. monotony
  1. ^ Wilkinson, Leland (23 April 2008). "Scagnostics". Retrieved 25 March 2022. {{cite journal}}: Cite journal requires |journal= (help)
  2. ^ Wilkinson, Leland; Anand, Anushka; Grossman, Robert (2005). "Graph-theoretic scagnostics". In Proc. 2005 IEEE Symp. On Information Visualization (INFOVIS: 157–164. CiteSeerX 10.1.1.329.1315.