Many social scientists view statistical techniques as
difficult. Often, this is attributed to insufficient statistical skills of
those scientists. However, we should also recognize that statistical techniques
are becoming increasingly complex. The question whether such complex techniques
are really needed is hardly given serious consideration: Their very complexity
seems to guarantee their statistical adequacy. However, complex techniques need
not be more adequate than simple techniques. Moreover, using complex techniques
without fully understanding them, one may miss important features of the data,
which one would notice using simpler techniques.
Our first purpose is to study how scientists actually use current statistical techniques. Because mental capacities and attitudes play a role here, collaboration with cognitive and social psychologists is needed. The next challenge is to design more usable, for instance simpler, techniques. These will be compared to existing techniques in terms of statistical adequacy (“Do they correctly summarize data?”) and practical applicability (“Do they provide comprehensible answers to research questions?”). The statistical adequacy will be assessed theoretically and by simulation studies, where data with a known structure are generated, and techniques are compared in terms of their capability to recover that structure. The practical applicability will be assessed by means of application to empirical data. Here, in particular, it will be studied whether researchers can effectively use the techniques to answer their research questions. Special attention will be given to the new problems posed in the analysis of the increasingly prevalent multiway data (e.g., scores of several subjects on several variables at several occasions). In our intended contributions to multiway techniques, again usability is of major importance.
For new techniques developed here, software will be prepared and made widely available to scientists. For new insights in the use of existing techniques, it will be suggested how to implement these in existing software.