ABSTRACT

The leading candidates for choice of expression are different for counted fractions than for amounts. The principal purposes which choice of expression can serve are approaching more closely constancy of variability; approaching more closely additivity of effect; approaching more closely linearity of response. For each of the goals, some progress can often be made by a choice of expression, thus dealing with removable aspects of inhomogeneous variance, of interaction, or of non-linearity, but often leaving to be faced aspects of inhomogeneous variances, of interaction, of non-linearity, that cannot be removed by choice of expression. The chapter is concerned with what expression would leave ‘some kind of variability’ more nearly constant. This is a related inquiry to having effects additive because variability can be thought of as the collective action of a pool of changes usually in factors with small effects.