ABSTRACT
In this chapter, the authors use the data to illustrate two main points. One is the use of general considerations to choose between alternative parametric regression relations. The other is the examination of the form of the random variation about such a relation. Elaborate model fitting would be unnecessary for the analysis of these data in isolation. In particular, many different parametric representations of the systematic variations are consistent with the data. If the random contribution to the ith observation is εi, simplicity of interpretation and fitting, and inspection of the data, suggest a multiplicative combination; that is, they consider an interpretation in which the proportional variation around the mean has the same form for all x. It is then of interest to compare that distribution with the exponential distribution, the Poisson process, and partly because use of the exponential distribution much simplifies more detailed analysis and consistency with exponential form is of general interest.
