Comparisons of Multidimensional Visualization Methods

• Activity Number: 43
• Type: Contributed
• Date/Time: Sunday, August 11, 2002 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistical Graphics*
• Author(s): Ted Mihalisin*+
• Affiliation(s): Temple University
• Keywords: multidimensional visualization ; nested dimensions ; very large datasets

Abstract

Three-multidimensional visualization methods utilizing nested dimensions namely; trellis-like displays, mosaic plots and TempleMVV graphs are discussed and compared in respect to the insights they provide and their performance. These techniques are applicable when the number of dimensions is no larger than ten to twenty. Only mosaic plots and TempleMVV graphs can be applied to datasets with large numbers of data points (millions or more). Trellis-like displays, consisting of collections of scatter plots of two continuous variables, conditioned by values of other variables, cannot be used effectively for large datasets because of both performance and over-striking problems, and the limitations of plotting "raw" data instead of statistical information. Although mosaic plots can be used to represent very large datasets, they can only display the sums of a response variable (or of count) not the means or other statistics. This limitation and a variety of display problems limit the types of insights that mosaic plots can provide. In order to make valid comparisons, all three methods are applied to the same dataset--namely, a census dataset involving 55, 000 persons.