the best way is to calibrate an analyzer with another method, like TRL, which doesn't depend on O,S,L and then measure them.

Alternatively, calibrate with an OSL, then add an airline, then measure the load. The peaks of the ripple are 6 dB above the load return loss (assuming the airline is perfect). Measure the Open and put data->mem and then measure the short. The difference is a measure of how far off the open and short are from the expected values. Since the short offset is almost always only delay, you can use this to model the open capacitance. One way is to change the capacitance, recalibrate, remeasure the ripple and see if it is better or worse. when you find the value that makes the ripple go to zero, you have the value that give the correct capacitance of the open. This depends on the airline being perfect, and the load being very good.

Alternatively, calibrate with an OSL, then add an airline, then measure the load. The peaks of the ripple are 6 dB above the load return loss (assuming the airline is perfect). Measure the Open and put data->mem and then measure the short. The difference is a measure of how far off the open and short are from the expected values. Since the short offset is almost always only delay, you can use this to model the open capacitance. One way is to change the capacitance, recalibrate, remeasure the ripple and see if it is better or worse. when you find the value that makes the ripple go to zero, you have the value that give the correct capacitance of the open. This depends on the airline being perfect, and the load being very good.