Hierarchical visualization techniques: a case study in the domain of meta-analysis
Abstract
Meta-analysis is a probabilistic technique which groups results from several studies addressing the same topic and produces a result that summarizes the whole. Results generated are graphically displayed without providing interactivity with the user or reproducing a friendly, easy to comprehend interface. In order to obtain a visual exploratory analysis of the most satisfactory results there are Information Visualization techniques which can be applied to map data graphically aiming to broaden the user cognition. This paper presents an analysis of hierarchical information visualization techniques to determine which of them can best represent a data structure, develops meta-analysis and applies information visualization techniques, obtained from the analysis, to the meta-analysis results obtained through the Software R.
Keywords
Download Options
Introduction
When data is represented in larger and more complex information systems, it is advisable to use some graphic representation. Some authors[1-2]define information visualization (IV) as a science based on the proposal of transmitting some message via graphic elements, improving understanding and facilitating the system maintenance. Visualization techniques (VT) are tools which enable the application of studies developed in IV, with their own characteristics, which allow the presentation of data from differentiated approaches[1].
Neto40 divides VTs into four main groups: geometric, pixel-oriented, iconographic and hierarchical. An evaluation of the techniques in all groups was carried out by Luzzardi[1], employing attributes such as: nature of the domain, data structure, type of information they best represent, type of data, size of domain, operations regarding data and operations regarding data representation. The focus of that study was to compare the techniques usability and not identify the best technique to represent a data structure.
VTs can be used to represent books in libraries, archive directories, browsers and statistical data, among others. One of the areas in which statistical analysis is employed is the meta-analysis.
Meta-analysis is a probabilistic technique which uses the combination of results obtained in several studies and produces results that summarize the data set. This technique can be applied to a fixed or random effects model through the existing software such as the Software R, which is open source free software[10].
The graphs generated by the Software R neither provide interactivity with the user nor reproduce a friendly and easy to comprehend interface. Thus, VTs are important tools to map data resulting from a meta-analysis into a graphic format aiming to broaden the user cognition[11].
Evaluation and classification of information visualization techniques are performed in Shneiderman[38] and Wiss et al[41] requiring user interaction and visual representation attributes on the screen. Luzzardi[1] proposes an evaluation of information visualization techniques based on visual representation attributes and interaction mechanisms.
As there is a large number of existing visualization techniques, a small group was selected for this study: hierarchical visualization techniques (HVT). This paper proposes a qualitative analysis of 18 hierarchical visualization techniques which have been reported in the literature using the criteria presented by Luzzardi[1] and other specific ones which allow the identification of the best hierarchical technique to represent data structure.
Conclusion
This article analyzed 18 hierarchical visualization techniques to create an evaluation process to identify the best technique to be applied to any data structure that can be hierarchically represented.
This process was used to elect the ideal visualization technique which could best represent the Funnel Plot graph, generated by the Software R.
Data plotted in the Funnel Plot graph was the result of meta-analysis carried out to identify the efficacy of the use of the fungicide Fluquinconazole to inhibit soybean rust disease.
The implementation of Bifocal Tree technique to represent the graph data Funnel Plot, developed by Software Gephi showed that the chosen technique could better represent the data structure of fluquinconazole database.
The exploitation of dynamic features offered by the technique resulted in a greater visual exploration, better and faster comprehension of the relevant data, and finally, the analysis of data provided further exploration of the data, and consequently increased the user cognition