Eric,
Do you not mean--
The roll axis is the imaginary line through the front and rear roll centres.
It is this imaginary line which some people believe should be parallel or near parallel to the gradient of the mass centroid. (another imaginary line)
It is the relationship between these two imaginary lines which govern mass transfer in roll. (the roll couple) Another controlling factor is the transverse shift of the Roll centres as the car rolls.
Roll stiffness as mentioned by Bob P dictates the amount of mass transfer and how it is distributed.
This is the " black art" part assisted these days by 3D computer simulation.
Yes, I do mean that as well. Some engineers would argue that the gradient of the mass centroid as a roll axis is false, but it is still the basis of many suspension designs. I happen to agree with the concept.
As a rule of thumb the place to start with a clean sheet is to keep the roll couple equal lengths in front and rear. In fine tuning it generally changes depending on the vehicles functional envelope, and multiple other factors of the realities of individual vehicle dynamics. There is no perfect formula for where to put the roll centers initially, its usually an educated guess based on the vehicle parameters.
the amount of weight transfer in the vehicle is not exactly controlled by spring rate, the mass will shift based on CG position, the springs are reactionary devices, not preventitive.
I work with many very sophisticated design level suspension CAD/CAE programs. They are very useful in initial phases, but vehicles always go through many changes when they hit the road as mules, and many unpredictable dynamics evolve.
Finding the final relationship between design type, geometry, alignment, spring-damper-swaybar, and ackerman settings is still very much an art that every manufacturer prides itself in having.
I am including an image of a simple mid-engine mass centroid and roll axis diagram so that it may help illustrate the concept.
Also, I have design level experience with the C5, and measured a C5 suspension with a 3 axis measuring system on a surface plate in the past, and I include images of it (notice the amount of front anti dive and the huge difference in the front vs rear arm orientation)
Also included is the simple CAD arm and steering arm position at ride height.
Cheers
Eric