ABSTRACT
Successive multiple regressions, as used to turn coordinates robustly orthogonal seems likely to absorb larger and larger shares of the original variability. The analysis of variance structure is the same for all new coordinates. It is reasonable to ask what scaling of the successive turned coordinates will make the comparisons of the successive analyses of variance useful. A major reason for working inside maturity classes is that fixing a year and a place does not provide the same environment for genotypes that mature after different periods of growth. The chapter discusses that scaling before turning would give equal spread for the various attributes, and, tacitly, that this was for statistical reasons.
