ABSTRACT
This chapter develops some of the broader strategical ideas involved in applying statistical techniques. They bear also on the interplay between statistical analysis and subject-matter interpretation. The scientific or technological interpretation of investigations never proceeds in isolation, although individual investigations may have very considerable, even if ultimately temporary, importance. An emphasis in statistical work has been on special theoretical stochastic models representing simplified theories of the system under investigation. Their properties can be studied either mathematically or by computer simulation. The use of these in statistical analysis is an attempt to relate data to some underlying physical or biological mechanism. In relatively complicated problems, as typified by regression problems with many explanatory variables, the main objective is an economical description of the data in a form consistent with external information. A particularly difficult issue is to decide when simple descriptive analysis is adequate and when explicit calculations of uncertainty involving probabilistic arguments are called for.
