Understanding of effects of multi-attributes sorted visualizations

Inkyoung Hur, Purdue University

Abstract

A table is a ubiquitous representation for multiple attributes, and sorting is one of the most widely-used interaction techniques in addition to a table in many software applications, such as spreadsheets. However, sorting appears to have limitations when used in analyzing multi-attribute data in a table. When a table is sorted by columns, it rearranges the whole table, so the insights gained from the previous arrangements are often difficult to be retained. Various multi-attribute visualization techniques (e.g., Table Lens and Parallel Bargrams) could solve the issue, but they could be too unfamiliar to the spreadsheet users. Thus, I proposed another set of multivariate visualizations, called “SimulSort” and “ParallelTable,” which are more familiar but overcome the issue. Extensive empirical studies were conducted to help understand how the proposed visualizations affect the performance in various tasks including analytic tasks and a multivariate decision-making task. A series of incentivized laboratory experiments helped quantify the impact of the four different techniques (i.e., a static table, a table with the sorting feature, SimulSort, and ParallelTable) on the performances of these tasks. Results showed that each visualization technique has advantages and disadvantages for completing different tasks. Experimental results showed the potential of the novel visualization techniques in promoting high-quality decision-making though not all the novel techniques are effective in supporting more primitive analytic tasks. The details of experimental results and implications are discussed.

Degree

M.S.I.E.

Advisors

Yi, Purdue University.

Subject Area

Industrial engineering

COinS