A plot suggested in 1972 by the American David Andrews as an alternative method to Chernoff faces for representing multivariate data in two dimensions. The plot can help to identify outliers and to establish similarities within groups of data items. For an item with values given by the row vector (a b c d e…), the function x(t) is defined as
The resulting graph is drawn for (−π≤t≤π), for each data item. The form of the graph is dependent upon the ordering of the characteristics a, b,…; the usual advice is to order the characteristics in declining order of their (supposed) importance.