A statistical user interface is an interface between a human user and a statistical software package. Whenever we use a statistical software package we want to solve a specific statistical problem. But very often at first it is necessary to learn specific things about the software package.
Everyone of us knows about the ''religious wars'' concerning the question which statistical software package/method is the best for a certain task; see [14] and [4] and related internet discussions. Experienced statisticians use a bunch of different statistical software packages rather than a single one; although all of the major companies (at least the marketing departments) tell us that we only need their software package.
Why do these wars, not only concerning statistical software packages, evolve? One of the reasons is that we need time to learn about the software package besides learning the statistical methodology. And the more time we need to learn to use the software package, the more we are defending ''our'' software package. But if we need to spend a lot of time for learning to use a statistical software package, then the question, whether this software package really has a good user interface, arises?
The basic problem is that the development of statistical software is started by experts of statistical methodology. Since they have to have a deep inside in their special fields, most of them have a very limited view to problems of other users. We generally do not consider the sex of the users, the ethnic or cultural background, the statistical knowledge or other factors when we create a user interface.
Thus the important questions we have to answer when we design a user interface are: What does the user want to do with this software package? And how can he do it effectively?
Fortunately, during years of development of software packages, we have
collected a lot of experience about human behavior and specific
differences because of sex, ethnic or cultural background and so
on. In the book of [24] a lot of rules have been
collected which should help the software developers to avoid the worst
problems. But the best way for developing a user interface is
a development cycle of designing, testing, redesigning, testing,
redesigning, 
This will take a lot of time and money, but
redesigning the basic components of a software package at late
development will be much more expensive or even impossible.
In this chapter only a subset of all statistical software packages, namely DataDesk 6.0, GGobi 0.99, R 1.7.1, SPSS 11.0 (English version), SYSTAT 10, XploRe 4.6 and Mathematica 4.3 will be used for illustrating purposes (see also the section ''Web references''). It covers a wide range of different statistical software packages.
In all statistical software packages we can find errors in the user interface design. User interface design is not a exact science as statistics, but it relies heavily on the way how we perceive information and how we react to it. That includes that in every design we will make errors before we can learn to avoid them in future. Therefore a lot of theories are available, partially explanatory, partially predicting, which should help us to design user interfaces.