The procedure would be:
Measure the S-parameters of your standards.
Make a circuit model for the cal standard and fit the simulated S-parameters to measured S-parameters.
Then mathematically derive the coefficients by doing a curve fit.
So your accuracy of the coefficients will depend on how accurately you measure the S-parameters. This in part will depend on the high quality cal kit you use. There are various levels of cal kits. Cheap and dirty is open/short/load. Next level is a high quality airline sliding load kit. Then TRL, which only depends on airline impedance. Nowadays you can throw in electronic cal kits like Keysight eCal, which may be the best of both worlds, fast and accurate.
Here are a few references from Keysight:
"Specifying Calibration Standards and Kits for Keysight Vector Network Analyzers"
The appendices have derivations of the cal coefficient models.
http://literature.cdn.keysight.com/litweb/pdf/5989-4840EN.pdf"Electronic vs. Mechanical Calibration Kits: Calibration Methods and Accuracy"
http://literature.cdn.keysight.com/litweb/pdf/5988-9477EN.pdf