I've extracted εr and tanδ from some dielectrics , on waveguide environment.
Firstly I've performed very accurate spar measurement with a VNA and then I've put spar into a linear simulator (Genesys®).
Into the simulator I've created 2 circuits, the 1st was a network with the same measured spar, the 2nd was the physical model (waveguide, dielectric sample).
Into the model, the εr and tanδ was variable.
Then I've adjusted the variables until the 2 spectral traces became overlapped.
Accuracy:
As usual for all measurements , the most important keyword is the accuracy.
Given the VNA accuracy, human errors ( cable movements etc.) a typical estimation of accuracy is U(εr) = +/- 2% ( for 1.1<εr<3) U(tanδ) = +/- 50% (for 1E-4<tanδ<3E-3).
If You are interested contact me privately.