Well, aside from the lowest mu types, you'd never use ferrite toroids for energy storage; and shapes are almost always used with air gap. Especially spool and rod types, of course. Energy storage goes as Bsat^2 / (2 mu_0 mu_r), so the high permeability makes for terrible inductors. (Even so, I've measured some shockingly high Qs on one modern ferrite, with minimal airgap. You'd still gap it of course, because of energy storage. And the reduced tempco is nice.)
More literature to be found on
https://www.fair-rite.com/ for example, including the impedance curves of many bead and ring cores. The complex mu plots (mu = mu' + j mu'') for the material basically give you the Q factor vs. frequency, assuming an ideal winding (and the same core..). Most roll off by 100kHz to 10MHz (roughly inverse with mu_r), that is to say, mu'' dominates at high frequencies; and complex inductance is just another way to say resistance.
But do mind that those material curves are themselves made on a typical ring core, and geometry plays a huge role in their shape: compare the impedance of say #43 beads, of various aspect ratios, and to rings, and to the material characteristic (which is measured on a... something around a T80 size I think?).
We can explain the diversity of these impedance curves by understanding wave effects: namely that wave velocity is very much slowed in ferrite, because index of refraction goes as 1/sqrt(mu_r e_r), and if mu_r ~ 1000 and e_r ~ 10, that's a good 0.01c propagation velocity, and it doesn't take much to start seeing 1/4 or 1/2 wave reflections (hence peaks or inflection points in the impedance curve) due to the length, diameter or thickness of these parts. Say 100MHz is 3m wavelength, 1/4 wave is 750mm, and 1/100th of that is 7.5mm, a typical scale for beads-on-leads, which have impedance peaks round about this frequency.
So it's all a bit hand-wavey up around or above the material cutoff frequency; and, ferrite still has some conductivity so exhibits eddy currents and skin effect (mainly a concern for high power levels (big transformers, like cores 10cm across) at power switching frequencies (~100kHz)). (And, hysteresis loss causes the same phase shift and absorption to the field, so acts equivalently to conductivity for purposes of skin effect.)
But we don't mind those effects for board-level components, and most importantly of all, they largely go away when an airgap is introduced, which seems to act to reduce mu_r (that is, if we were to assume the core were solid, what its effective mu_r would be), and the introduced amount is lossless (air!) so greatly increases the Q factor; even mu_r ~ 10k ferrites are usable at 100s kHz with an airgap, since while they might be dissipating relatively high core loss, the reactive power at the winding is just that much more (and Q = VAR / P).
(Not to be confused with S = P + jQ, the symbols for apparent, real and reactive power used in power electronics. Since we're using Q for "quality factor", I'll just quietly replace that with VAR (volt-amps reactive).)
Tim