The following note was created by Jim Pasha.
I did some calculations some years ago and the following factors apply to the rate. The lengths and diameter of the bar are the largest contributors. The material is the next largest contributor.
Considering the material used, diameter and length, this is what the diameters represent:
Length = 675 mm (26 5/16 in.)
diameter (mm) raw rate (lb/in) effective rate (lb/in) original application 22 464 292 Stock 924 23 554 344 Stock 944 23.5 628 379 VW Transporter 25 773 486 Stock 944 turbo, 944S2 and 968 26 904 568 27 1062 661 28 1212 765 29 1400 886 30 1603 1008
This is only part of the equation. The figures are only for twist of the bar itself. If you add the spring plate, you have to factor in the total distance of the rotation, center of the torsion bar to canter of the rear wheel. That is the effective lever length. As the lever arm gets longer, the spring rate decreases. The lever, spring plate in this case, will factor into the equation and reduce effective rate.
One size torsion I did mention was the 23.5 mm VW transporter torsion bar which I found to work very well at the rear of some early 924s I've owned. This was something we did in early 1977 to greatly improve rear suspension spring rate. It was very cheap to do though took some time to make the changes. As the 924 family rear torsion bars are based on VW units, any of those will work at the rear of these cars.
I think the big thing was the use coil springs allowed for easier changes and progressive rate springs. A major disadvantage is that torsion bars, unlike coil springs, usually cannot provide a progressive spring rate, forcing a compromise between ride quality and handling ability - progressive torsion bars are available, but at the expense of durability since they have a tendency to crack where the diameter of the bar changes. The major objective was improved ride quality with the expected Porsche handling.