Department of Statistics, UCLA

In the situations typically encountered in the social sciences the methodology of traditional statistics is neither a good description of data analysis practice nor a good prescription to arrive at satisfactory summaries of the data. On the other hand the traditional statistical techniques are excellent data analysis tools, and statistical models are useful devices that can be used to develop and evaluate data analysis techniques. These general considerations are applied in this paper to evaluate the usefulness of techniques such as multidimensional scaling and correspondence analysis.