A Visualization Tool for Learning Statistical Analysis in Multi Tabular Datasets

Abstract

The ability of the human mind to perceive visual informationmakes visualization not only useful, but a powerful toolfor information discovery. Answering questions about complexrelationships requires the analyst to choose a statistical analysistechnique that makes relationships visually discernible. Oftenthe proper technique is dependent on the characteristics of thedataset, such as dependency among variables, sample size, andtypes of data (ordinal or categorical). In this work, we proposea web based interface approach that visualizes various statisticaltests and displays the distributions of data using color codingschemes. With our system, a user can select multiple variablesinteractively, and the resulting selections will be visualized tohelp the user understand the data and statistical formulas used toshow it. This capability allows a user to quickly evaluate differentsubsets of a large, complex dataset for statistical correlations. Tovalidate our approach, we performed a controlled user study toevaluate the ease of use of our system, and to test the effectivenessof our interface. We see our system as directly applicable to dataanalytical tasks, as well as a useful teaching tool for those learningdata analytics.