Types of GUIs

The concept of GUI is general. It just means providing an event-driven user interface managed with the mouse (the most classical graphical interfacing device). Usually, a GUI is based on menus and dialog boxes. Regarding the data analysis, there are different ways to organize the commands and to display both the data and the results. Here is a typology of the different GUIs that have been used so far in data analysis:

• MDB (menus/dialog boxes): this GUI is commonly found in current statistical packages like Minitab, SPSS, Systat, Statistica,... Splus proposes also a MDB...

• Spreadsheet: like Excel, Gnumeric... The major part of the interface is a table constituted of cells. Each cell contains numbers, text or formulas. The computation is done by recalculating the formulas... Some spreadsheet packages allow extensions (addins, plugins,...) to add advanced data analysis methods to the panels of formulas available.

• Notebook: the notebook interface is an extension of the word processor. The document mixes text and images, like in a classical word processor. In addition, paragraphs can contain code that is executable by the calculation engine and some of the images are indeed graphs generated by this code. A typical notebook interface is Mathematica. The notebook interface can also be viewed like a live math scratch book (the metaphor preferred by Mathcad, another notebook-based program). It can also be considered as an extension of the command line interface, where not only text results are displayed under the command, but also graphs, and where rich-formatted text can be included in between.
• Web-based: the web-based interface presents active web pages with forms that are filled and trigger analyses on a server. Its main advantages are the Client/Server architecture, but it can work locally also.

Which is the best GUI?

One GUI cannot be said to be better than another, nor a GUI is better than the CLI. The answer depends upon the context, that is, the environment, the analysis and the user. Giving a general data analysis software should adapt to the context, the best user interface is perhaps the most versatile one. It should offer various ways to drive the analyses: CLI, MDB, spreadsheet or notebook, depending on the user's needs and preferences.

The environment, mainly the platform imposes some habits to the user. For instance, most Unix/Linux users are accustomed to a CLI, while Windows/MacOS users are prone to prefer a GUI with the same look and feel as their preferred operating system. The way the data analysis package is used is also a major criterion. Whether for teaching, for research or for demonstration... Finally, the user skill is key aspect. A beginner prefers a point and click user interface, while the expert likes to get full control of the process through a CLI, and consequently, dislikes rigid and limited GUIs. This latter aspect is controversial. Many teachers prefer to use a CLI, even with their beginner students, because it forces good habits in data analysis (the user must really know what he is doing). On the contrary, a beginner in front of a simple GUI can click everywhere and get results anyway, even if these results are meaningless.