What are the common causes of measurement errors when using a vector network analyzer?

The common causes of measurement errors when using a Vector Network Analyzer (VNA) are: inadequate or expired calibration (the most common error; the VNA's error correction depends on a valid calibration performed at the test ports using known calibration standards; using the VNA without calibration or with a calibration shifted by temperature, frequency range, or cable disturbance introduces errors of 1-10+ dB in gain/loss and 10-30+ degrees in phase), cable movement after calibration (flexing the test cables after calibration invalidates the error correction because the cable's phase and loss change; use phase-stable cables or re-calibrate after cable movement), calibration kit mismatch (using a calibration kit that does not match the connector type or not entering the correct cal kit model/coefficients results in systematic errors), incorrect reference plane (the calibration standard defines the measurement reference plane; if adapters or fixtures are added between the calibration reference plane and the DUT, the measurement includes the adapter's S-parameters), source power too high or too low (measuring a nonlinear device at too high a power gives compressed S-parameters; measuring at too low a power reduces SNR and increases noise-related uncertainty), IF bandwidth too wide (wider IF BW increases measurement speed but reduces dynamic range and increases noise; conversely, too-narrow IF BW slows measurement unnecessarily), and port extension errors (using port extension or de-embedding with incorrect parameters shifts the reference plane incorrectly).