ABSTRACT
The previous chapter illustrated the poor correspondence between mul tivariate Normal models on the one hand, and complex bathtub /-type functions as well as rough-edged population pyramids on the other. Ageadjustment procedures that were devised to deal with these complex curves have a long history. However, during much of this period, they have been applied exclusively to deal with a single binary-valued response variate that assumes only one of two values of the sort dealt with by logistic regression.
